JPH0990026A - Object detecting device and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately track an object. SOLUTION: The detected object which is the closest to the predicted position PE of an object 53P moving at a known speed from a known position PP after a fixed period of time is not simply recognized as the object 53P (stored object), but the object at such a position that the value obtained by subtracting the distances (absolute values of displacement vectors Pd) between the predicted position PE and positions of detected objects 54, 55, and 56 from the inner products of the predicted position PE in the direction of a speed vector starting from the position PP and the position vectors Pi ending at the positions of the objects 54, 55, and 56 becomes the maximum is decided as the object 53P after the prescribed period of time. Therefore, objects can be tracked with higher accuracy.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is mounted on an automobile, for example, and detects relatively moving objects such as a plurality of preceding vehicles (front vehicles) or oncoming vehicles with respect to the automobile,
The present invention relates to an object detection apparatus and method capable of suitably tracking (tracking) a detected object.

In this object detecting apparatus, a so-called ID (Identifier) is added to a detected object.
to keep track of.

[0003]

2. Description of the Related Art First, a general tracking principle using an object detection device will be described. That is, a moving object whose position (think of it as a past position) has been detected up to a certain point by some method is called a memory object, and a moving object whose position (think of it as a current position) has been detected by a certain method at the next point of time (the above-mentioned It may be the same as or different from the moving object.) Is called a detected object. In this case, the same moving object can be tracked by taking the correspondence between the stored object and the detected object at each time point.

As a conventional technique based on such a tracking principle, Japanese Patent Application Laid-Open No. 6-150195 or Japanese Patent Application Laid-Open No. 6-195195.
The "object detection device" disclosed in Japanese Patent Publication No. 150196 can be mentioned.

This object detection device predicts the position where the storage object exists at the next time point from the position information and the velocity information of the storage object at a certain time point, and this predicted position {this predicted position is a past position. Is given as a position obtained by adding the moving distance (the speed information × time taken) to. } And the detection position of the detection object detected at the next time point, and adopts a technique of identifying the detection object having the smallest distance difference between the predicted position and the detection position as the storage object. There is.

[0006]

However, in the above-mentioned conventional object detecting apparatus, since the predicted position and the detected position, which are the objects of comparison, are scalar quantities, it causes an erroneous correspondence described below. The problem occurs.

As shown in FIG. 12A, an object detection device such as a stereo camera (not shown) is mounted, and a detected object ID is detected within the moving object detection range 4 of the vehicle 3 traveling in the direction of arrow M.
When the storage object 1 having the number 1 and the storage object 2 having the ID number 2 exist, when the lanes are changed to the positions surrounded by the dotted line and the left and right positions are switched by the time when the next position is detected, FIG. , The storage object 1 with the ID number 1 is recognized as the storage object 2 with the ID number 2, and the storage object 2 with the ID number 2 is recognized as the storage object 1 with the ID number 1.
So-called mishandling (misrecognition), in other words, so-called wrong ID succession may occur.

Similarly, as shown in FIG. 13A, the storage object 1 with ID number 1 and the storage object 2 with ID number 2 have moved relative to each other by the dotted line until the next position detection time. 13B, the storage object 1 with ID number 1 is recognized as the storage object 2 with ID number 2 and the storage object 2 with ID number 2 is stored with the storage object with ID number 1 as shown in FIG. 13B. It may be recognized as 1.

Further, as shown in FIG. 14A, when the storage object 1 and the storage object 2 are ranked, and the storage object having a higher rank (in this case, the storage object 1) is associated (searched), When the storage object 2 moves to a position surrounded by a dotted line and is aligned with the storage object 1 having a high order in a straight line in the traveling direction M, the storage object 1 having a high order can be temporarily detected due to the influence of noise or the like. 14B, the storage object 2 of the next rank is replaced with the storage object 1 of the higher rank as shown in FIG. 14B.
May be identified as

The present invention has been made in view of the above problems, and an object thereof is to provide an object detecting apparatus and method capable of accurately tracking a plurality of moving objects without erroneous correspondence. To do.

[0011]

According to the present invention, an object position detecting means for detecting a current position of a relatively moving object at regular time intervals, a current position and a past position of a detected object, and the constant time. Speed vector calculation means for calculating the speed vector of the detection object based on, storage means for storing the current position and speed vector of the detection object as a pair,
Based on the current position and velocity vector of the stored detected object (hereinafter, referred to as a memory object), the current position is set as a reference past position, and the reference past position is set as a starting point, and the storage object after the lapse of the predetermined time. Prediction means for calculating a predicted position vector having the predicted arrival position as the end point, and a position vector having the current position of the object newly detected by the object position detection means as the end point and the reference past position of the storage object as the start point Of the calculated position vector,
And a correspondence determining means for identifying a position vector corresponding to the predicted position vector as a position vector representing the current position of the storage object.

According to the present invention, the predicted position vector after a fixed time is calculated with the position of the storage object as a reference (start point), while the position of the detected object detected at the present time is set as the end point and the position of the storage object is set as the start point. Then, a position vector (which is a candidate position vector because it is a candidate position vector for obtaining the current position of the storage object) is calculated. Then, the correspondence determining means identifies the candidate position vector corresponding to the predicted position vector as the position vector representing the current position of the storage object.

In this case, the current position represented by the tip of the identified position vector and the velocity vector calculated based on the identified position vector and the fixed time are paired,
A moving object can be continuously tracked by providing a memory updating means for updating the contents of the record of the memory object.

Further, the storage updating means deletes the record of the storage object, in which the corresponding identified position vector does not exist a predetermined number of times, in the correspondence determining means, and the corresponding identified position vector is stored. By storing the current position and velocity vector of a newly detected nonexistent object as a new record in the storage means as a new record, it is possible to reliably track the moving object and to efficiently track the moving object. You can

Further, the correspondence determining means has Pe as the predicted position vector, Pi as the position vector having the current position of the object newly detected by the object position detecting means as the end point and the reference past position of the storage object as the start point, and an arbitrary coefficient. Is k and the degree of correspondence is C
When i, the correspondence Ci is Ci = Pe · Pi−k | P
The detected object corresponding to the storage object is easily identified by identifying the position vector Pi represented by i-Pe | and the position vector Pi having the maximum value of the correspondence degree Ci as the position vector corresponding to the predicted position vector Pe. You can

The reason is that the inner product Pe · Pi of the vector is maximum means that the position vector Pi and the predicted position vector Pe are in the same direction, and further, the displacement vector Pi
This is because that the minimum absolute value (scalar amount) of -Pe means that the speed difference and the direction difference are minimum.

Furthermore, the method of the present invention is an object detection method used when detecting and tracking an object that moves relatively, in which the object after a certain time has elapsed from the speed obtained at a known position of the object. Predicted position calculation process of calculating the predicted position, at a time point after the elapse of the certain time, an object position detection process of detecting the position of the object, a direction connecting the known position of the object and the predicted position of the object, A correspondence degree calculation step of calculating a displacement angle formed between a known position of the object and a direction connecting the detected position of the object, and calculating a distance between the predicted position of the object and the detected position of the object, The position of the detected object whose displacement angle is small and whose distance is small is determined to be the actual position of the object whose known position is known after a certain period of time. That.

According to this method, the detected object closest to the predicted position is not made to correspond to the object (memory object) whose position is already known, but the displacement angle from the velocity vector direction based on the position of the memory object is used. Since the degree of correspondence is evaluated in consideration of the distance between the predicted position and the detected position, the object can be tracked more accurately.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to the functional block diagram of FIG.

FIG. 1 shows the configuration of functional blocks according to the present invention of an object detection device 8 mounted on a vehicle (not shown) or the like.

In FIG. 1, the position detecting means 10 is a sensor capable of detecting the three-dimensional position or the two-dimensional position of a plurality of objects moving relative to the vehicle at fixed time intervals (sampling time). And, for example, two CCDs
A stereo camera using a video camera, a scanning type / array type laser radar, a millimeter wave radar, or the like can be used.

The storage means 11 is a RAM (random access memory), and stores, for example, the two-dimensional position or three-dimensional position, speed, and width of a moving object that is an obstacle to another vehicle, a person, or another vehicle. Information (hereinafter referred to as attribute information) is stored. In this embodiment, the attribute information is collectively stored for each storage object, and the attribute information is added or deleted as a unit. Hereinafter, the unit of this attribute information is called a record.

FIG. 2 shows an example of the structure of the record 51. As shown in FIG. 2, the record 51 defines the records before and after the record 51, that is, the pointer of the next record and the pointer of the previous record.

Further, the time of creation (time of calculation) of the record 51, in other words, the detection time t ₀ when the object is detected is stored. Furthermore, the position of an object in the detection time point t ₀ (x, y), also each time point t _-1 to the detection point t ₀ from the front a plurality of past several times, ... the object position of the t _-n (x , Y) are stored respectively. The sampling time, that is, the detection time interval, is a fixed time between the time points before and after, for example, time point t ₀ and time point t ₋₁ . It should be noted that the position (x, y) of the object is stored as a coordinate point having the vehicle position at the time of detection as the coordinate origin, that is, a position vector.

Further, by taking into account the positions of the objects at a plurality of consecutive time points, appropriate filtering processing is performed on the position information and time information of these objects to remove noise contained in the position information and to calculate the velocity. The accuracy can be increased. The time array may be, for example, about 10 to 15 arrays when the object detection cycle which is the sampling time is 33 ms. Based on such an idea, the velocity vector V calculated at the time t ₀ (this velocity vector V is also composed of the velocity component vector Vx in the x direction and the velocity component vector Vy in the y direction) is the record 51. Memorized in.

As will be described later, the corresponding counter in the record 51 is a counter that stores, as count values, the number of times the object represented by the record 51 is detected and the number of times it is lost (non-detection number).

FIG. 3 shows the structure of the list 52 of the record 51 for a predetermined storage object. Record 5
1 is created for each storage object. As can be seen from FIG.
The record 51 (think, for example, as record 2 in FIG. 3) is a record 51E (think as record 1 in FIG. 3) before and after by a pointer and a record 51P (FIG. 3).
I think it's record 3, inside. ) Is connected to the list 52. When a new record related to a new detection position is created, that record is added to the end of the list 52 shown in FIG. 3 and becomes the record 4.

Since the list 52 is created for each moving object, the order of the records remains the priority order for searching the moving object. Further, since any two records can be exchanged, it is easy to change the search priority.

In FIG. 1, the correspondence degree calculating means 12 which also functions as the velocity vector calculating means and the predicted position vector calculating means (prediction means) is newly detected by the position detecting means 10 as described below, for example. An object (detection object),
The degree of correspondence with the object (memory object) stored in the storage unit 11 is calculated.

FIG. 4 is a diagram for explaining the calculation of the degree of correspondence. FIG. 4 shows a code ◯ detected by the object detection device 8.
A storage object (also referred to as a storage object ◯) 53P, 53
An example of the positional relationship between currently detected detected objects (also referred to as detected objects Δ) 54, 55, and 56, which are denoted by P ′, 53P ″, ... And Δ, is shown. First, the past position and speed of the memory object ◯. To the next time (when the detection object Δ is detected, after all,
It is the current time. ) A predicted arrival position (also simply referred to as a predicted position) PE that is predicted to be detected in () is set as an end point, and a past position (also referred to as a reference past position) PP (a time point one sampling time before the present time point) is set. A predicted position vector (which can be regarded as a displacement vector, but is referred to as a position vector because it is a vector based on the past position PP) Pe is obtained.

Next, as shown in FIG. 4, the object search range 58 is determined around the predicted position PE. The search range 58 is the object detection range 4 shown in FIGS.
It is set within the range. The size and shape of the object search range 58 depends on the target object, and in this case, since the target object is a car, a person, or the like, it can be determined from the limit of the physical moving speed of those. . The shape may be a quadrangle, a circle, or any other shape. When the shape is a circle, the search range can be narrower than that of a quadrangle, so that the total calculation time can be reduced.

Next, as shown in FIG. 5, a position vector Pi whose starting point is the current position (the position shown in FIG. 4) of the detected objects 54, 55, 56 and whose starting point is the reference past position PP.
(Since the drawing becomes complicated, the position vector Pi starting from the reference past position PP and ending at the current position of the detected object 54
Only draws. This can also be regarded as a displacement vector. ) Respectively. Since the position of the end point of any one of these position vectors Pi is identified as the current position of the storage object 53P,
From this meaning, the position vector Pi is set to the candidate position vector P
It is also called i.

Next, the correspondence degree calculation means 12 calculates the storage object 53P and the detected object from the predicted position vector Pe and the candidate position vector Pi based on, for example, the calculation result of the calculation formula shown in the following formula (1). The correspondence degree Ci of each of 54, 55 and 56 is calculated.

Ci = Pe · Pi−k | Pi−Pe | = Pe · Pi−kPd (1) In the formula (1), the correspondence Ci is the inner product Pe · of the predicted position vector Pe and the candidate position vector Pi. Absolute value of displacement vector {(Pi-Pe) = Pd} obtained by subtracting predicted position vector Pe from candidate position vector Pi from Pi |
It is represented by a value obtained by subtracting the value obtained by multiplying Pi-Pe | by an arbitrary coefficient k. Note that the symbol i generally represents the candidate position vector Pi for the i-th detected object counting from 1, but in the case of the examples of FIGS. 4 and 5,
It is assumed that i = 54, 55, and 56 in the specification creation technique.

Then, the correspondence determining means 13 in FIG. 1 identifies the end position of the candidate position vector Pi having the largest value of the correspondence Ci as the current position of the storage object 53P.

In this case, the maximum value of the equation (1) is that the inner product Pe · Pi of the vector is the maximum value, in other words, the predicted position vector Pe and the candidate position vector Pi are in the same direction, and the displacement vector It is obtained when the absolute value | Pi-Pe | of Pd = (Pi-Pe) is the minimum value, that is, the speed difference and the direction difference are the minimum values.

FIG. 6 shows the search range 58 in the state of FIG.
Position vector P5 of the detected objects 54, 55, 56 in the
4, P55, P56 and the predicted position vector Pe (1)
The correspondences C ₅₄ , C ₅₅ , and C ₅₆ are three-dimensionally represented based on the calculation result of the equation. In FIG. 6, the Z axis of the coordinates represents the degree of correspondence, the X axis and the Y axis represent the position of the detected object, and the mesh represents the distribution of the degree of correspondence when the detected object exists at each position within the search range 58. ing.

From FIG. 6, it can be seen that the value of the correspondence C ₅₅ is the maximum. Therefore, the detected object 55 is the memory object 53P.
Can be easily identified.

In this case, according to the prior art, as shown in FIG.
, The position closest to the predicted position PE, or the reference past
The detected object 54 existing at the position closest to the position PP is stored.
Wrongly corresponded by mistake as a detected object corresponding to the object 53P
Possibility of causing it, in other words, actually a memory object
The detected object 54 corresponding to 54P corresponds to the memory object 53P
There is a possibility that it may cause a false correspondence (misrecognition) with a detected object that does
However, according to the present invention, as shown in FIG.
Correspondence C of body 55₅₅Is the correspondence C of the detected object 54 ₅₄than
The value is slightly high, and the detected object 55 is the memory object 53.
Properly succeed the ID as a detected object corresponding to P
It can be carried out.

The effect of the arbitrary coefficient k will be generally described. In the equation (1), when the value of the arbitrary coefficient k is set to a large value, the detected object existing at a position closer to the direction of the predicted position vector Pe is detected. When the correspondence Ci has a relatively large value and the correspondence is taken, so-called directivity becomes strong. By doing so, even when a plurality of obstacles approach each other, the probability of erroneously assigning an ID is reduced. On the contrary, when the value of the arbitrary coefficient k is reduced, the directivity becomes weak, and even if the detection position of the obstacle is not stable due to the influence of noise or the like, the correct ID can be continuously added without losing sight.

As described above, the correspondence determining means 13 in FIG. 1 identifies the end position of the candidate position vector Pi having the largest value of the correspondence Ci as the current position of the storage object 53P. However, to be exact, the memory object 53
The detected objects 54, 5 whose correspondence Ci is calculated for P
The one having the highest correspondence degree Ci is selected from among 5 and 56, and if the value is a value equal to or larger than a predetermined threshold value, it is determined to correspond to the storage object 53P.

In FIG. 1, the speed calculating means 14 is a storage object 5 in which the corresponding detecting object 55 is detected by the correspondence determining means 13.
3P, the past position of the storage object 53P {the reference past position PP and the past position P of the storage objects 53P ', 53P "...
.. (see FIG. 7)} and the detected object 55.
Using a suitable filter, for example, a Wiener filter or a Kalman filter, the velocity vector (predicted velocity vector or simply velocity) V at the current position 55N is calculated from the current position 55N. The velocity vector V is
For example, the velocity component Vx in the x-axis direction (note that this value will be considered as a scalar amount hereinafter for the sake of clarity). When considering the vector amount, the unit vector of the x-axis may be multiplied. .) And a velocity component Vy in the y-axis direction (also considered as a scalar amount).

In FIG. 1, the storage updating means 15 stores each storage object information sequentially from the head of the list 52 (see FIG. 3) stored in the storage means 11 based on the determination of the correspondence determining means 13. Record (here, record 51
E, 51, 51P). That is, if the storage object has a corresponding detection object, the current position of this detection object is set as the latest position of the storage object, and the velocity V, which is the vector amount calculated by the speed calculation means 14, is set as the storage object of the latest position. Speed. In addition, when the storage object does not have a corresponding detected object, and the count value of the corresponding counter in the record of the storage object is equal to or less than a predetermined threshold value, the record of the storage object is deleted, while When the count value of the corresponding counter is larger than the threshold value, for example, the current position is estimated based on the following equation (2) and the content of the record is updated.

Xi = Xi−1 + Vx × Δt Yi = Yi−1 + Vy × Δt (2) Here, Xi and Yi are the estimated current position and Xi−1 and Y.
i-1 is a past position, Vx and Vy are past velocities, and Δt is a sampling time of the position detecting means 10.

After updating all the contents of the record of the list of storage objects in this way, a record of a new storage object that does not correspond to the storage object is added to the end of the list.

The above description is an explanation of one embodiment of the present invention with reference to the functional block diagram of FIG.

Next, a more specific embodiment of the present invention will be described.

FIG. 8 shows the structure of a moving obstacle detection device 30 to which the object detection device according to this specific embodiment is applied. This moving obstacle detection device 30 calculates a position detection device 20 that obtains a set of relative position information of an object existing within the detection range from itself, and a set of position information output from this position detection device 20. And a computer 21 as an arithmetic and control unit (which also serves as a velocity vector calculation unit, a prediction unit, a correspondence determination unit, and a storage update unit) for processing and continuously outputting the position of the obstacle.
The position detection device 20 can use a stereo camera or the like for so-called binocular visual distance measurement. The stereo camera has CCD video sensors, which are left and right area sensors, and supplies pixel position information (collection of position information) of an object displayed in the image frames of both area sensors to the computer 21. With the computer 21, for example,
Match the images of both area sensors to identify the object, calculate the distance from itself (position where the stereo camera is located) to the object based on the principle of triangulation, and detect (position) the object position. You can

Therefore, the computer 21 receives the RA for input which receives the set of position information from the position detecting device 20.
M22, a CPU 23 that extracts a set of detected objects from the set of the position information and obtains their positions, and performs correspondence processing between the stored objects and the detected objects, and various programs such as a list of stored objects, a system program, and a control program. A RAM 24 for storing parameters and the like, a RAM 25 for outputting the updated position information of the storage object as a result of the calculation by the CPU 23, a bus 26 for mutually connecting these, and the like.

Next, the moving obstacle detecting device 30 shown in FIG.
The following object tracking operation operation will be described with reference to the flowcharts of FIGS. 9 to 11. The computer 21 is the control subject of processing execution.

In the general flow chart of FIG.
First, based on the set of position information output from the position detection device 20, the relative position of the object is calculated from the distance to the object and the pixel position, and the position information is clustered to extract the positions of the detected objects, A representative value of a set of position information belonging to a cluster is obtained and used as the position (current position) of the detected object. In this case, the representative value of the set of position information can be obtained, for example, by calculating the average of the elements of the set (step S100).

Next, the search pointer is set (set) to the head of the storage object list stored in the RAM 24 (step S200).

Then, it is determined whether or not the content of the search pointer has become NIL (nothing), that is, whether or not the storage object list has been examined to the end (step S30).
0).

If the determination in step S300 is negative, the pointer value of the next record in the records pointed to by the search pointer is set in the search pointer, and the correspondence determination process is performed (step S500). In this correspondence determination process, the correspondence Ci between the storage object stored in the record pointed to by the search pointer or each detection object within the search range is calculated to determine whether there is a detection object corresponding to the storage object. If so, in some cases it is determined which detected object corresponds.

If the determination in step S300 is affirmative, it is assumed that the storage object list has been exhausted, and the current speed is calculated from the determined current position of the detected object and the past position of the storage object (step). S600).

Next, the storage object list is examined from the beginning, the record of the storage object is updated according to conditions such as the presence or absence of the corresponding detected object, the value of the corresponding counter, and the detection of the absence of the corresponding storage object. The object is added to the list (step S700).

Finally, according to the contents of the updated storage object list, the position of the vehicle or the like which is an obstacle is written in the output RAM 25 together with the ID, so that the position information supplied from the position detecting device 20 at every sampling time is stored. The object detection process based on the set ends (step S800).

Details of the correspondence determining process in step S500 will be described below with reference to the flowchart of FIG.

First, the search range of the detected object is determined from the past position of the storage object, the velocity vector from the past position, and the predicted position based on the sampling time (step S501). Here, the size and shape of the search range can be determined from the limit of the physical movement speed of the detection target. For example, considering that an oncoming vehicle is detected on a toll road with one lane on each side, the relative speed is 250 km / h.
If the sampling time of the moving obstacle detection device 30 is 33 ms, if the range of a circle having a radius of 2.31 m, which is the moving range during one detection time interval, is set as the search range, the memory is stored. The detected object corresponding to the object can be almost detected.

Therefore, the detected objects within the search range thus determined are listed (step S50).
2).

Then, it is determined whether or not there is a detected object listed (step S503). When there is no detected object, the count value of the corresponding counter is decremented by 1 in order to store that there is no corresponding detected object (step SS505). When the count value of the correspondence counter becomes a value equal to or smaller than a predetermined threshold value, it is determined that the object disappears from the detection range of the detection device in the storage update processing step S700 described in detail later, and delete.

If a detected object is present in the determination processing of step S503, for the detected object listed in step S502, from the past position and speed of the storage object and the current position of the detected object, for example, The correspondence Ci is calculated based on the above-mentioned formula (1) (step S).
504).

Ci = Pe · Pi-k | Pi-Pe | (1) Here, Pe is the predicted position PE after the elapse of the sampling time in the velocity vector direction, starting from the past position of the memory object.
(See FIG. 5) The predicted position vector whose end point is (see FIG. 5), Pi is a candidate position vector whose start point is the past position of the storage object, and whose end point is the position of the detected object, and k is an arbitrary constant.

The x-axis direction component of the predicted position vector Pe is Pex (scalar amount), the y-axis direction component is Pey (scalar amount), and the x-axis direction component of the candidate position vector Pi is Pix (scalar amount). , Y-axis component is Piy
If (scalar amount), the equation (1) can be calculated as shown in the following equation (3).

Ci = PexPix + PeyPii−k√ {(Pix−Pex) ² + (Pii−Pey) ² } (3) From the equation (3), the number of operations (the RAM, the CPU, etc.) when calculating the correspondence degree Ci (Because the resource is used and the calculation time is required, it will be referred to as calculation cost hereinafter.) Is 8 multiplications, 3 additions, 1 subtraction, and 1 square root calculation.
This is not so different from the calculation cost (two multiplications, one addition, two subtractions, and one square root calculation) when the closest detected object is associated with the related art. Incidentally, in order to further reduce the calculation cost, the correspondence Ci can be obtained by the following equation (4) without square root calculation.

Ci = PexPix + PeyPyy-k {(Pix-Pex) ² + (Piy-Pey) ² } (4) Next, among the detected objects in the list, the one with the highest degree of correspondence Ci is associated with the storage object. A candidate to be selected is selected (step S506).

Next, it is determined whether or not the selected detected object has the correspondence Ci that is sufficient to determine that it corresponds to the storage object, based on whether the maximum correspondence Ci is a value equal to or larger than a predetermined threshold value. (Step S507). If the value is less than the threshold value, the process returns to step S505.

If the determination in step S507 is affirmative, the count value of the corresponding counter is incremented by 1 in order to store the fact that the detected object corresponding to the stored object was present (step S508).

In the final process of the correspondence determination process, the detected object corresponding to the stored object is deleted from the set of detected objects (step S509). By processing in this way, the order of the list of storage objects becomes the corresponding priority as it is, and there is no need to perform the processing when a plurality of storage objects compete for one detected object, which significantly reduces the calculation cost. Can be reduced. In this embodiment, as will be described later, a newly detected memory object is added to the end of the list. However, this is as it is, and an object tracked (tracked) for a long time has a higher priority. I mean.

Next, the details of the storage update process of step S700 will be described with reference to the flowchart of FIG.

In step S701, the search pointer is set to the head of the storage object list stored in the RAM 24. In step S702, if the content of the search pointer is NIL (nothing), it is determined that the storage object list has been searched to the end, and the process proceeds to step S708. Otherwise, in the record 51 pointed to by the search pointer (see FIG. 2). Set the value of the pointer of the next record to the search pointer,
It proceeds to step S703.

In step S703, if the stored object has a corresponding detected object, the process proceeds to step S704, and if not, the process proceeds to step S705.

In step S704, the past position of the record 51 of the storage object and the oldest information within the past time are deleted. Then, the current position of the corresponding detected object is stored in the array of past positions, and the detection time (detection time, detection time)
Is stored in the time array, and the speed obtained by the speed calculation process in step S600 is stored in the speed storage area of the record 51 (FIG. 2).
Medium, velocity vector V (Vx, Vy) column}.

In step S705, the process is decided according to the value of the corresponding counter. If the counter value is less than or equal to the predetermined threshold value, the process proceeds to step S706, and if not, the process proceeds to step S707.

In step S706, the object is erased from the detection range of the moving obstacle detection device 30 or the detection range of the position detection device 20 (that is, the moving object detection range 4 shown in FIGS. 12 to 14) and stored. Delete the object record.

In step S707, it is assumed that the object has disappeared temporarily and the current estimated position is obtained from the past position and velocity by the above equation (2), and the past velocity value is set as the current velocity value.

In step S708, it is checked whether or not the set of detected objects is an empty set. If it is not an empty set, the process proceeds to step S709, and if it is an empty set, the process ends.

In step S709, as many records as storage objects are generated, the current position of the detection object is stored in the past position array of the record, and the detection time point is stored in the time array. . The velocity value and the count value of the corresponding counter are set to zero values and added to the end of the storage object list.

As described above, according to the above-described embodiment, the detected object closest to the predicted position is not made to correspond to the stored object, but the displacement angle from the velocity vector direction based on the past position is taken into consideration. Since the degree of correspondence is evaluated by inserting it, the effect that the object can be tracked more accurately is achieved.

The degree of correspondence Ci is not obtained by directly using the inner product as shown in the equation (1) = (3) or (4), but by using an equation obtained by inversely calculating the inner product. You can also

That is, the degree of correspondence Ci can also be calculated by the equation (5).

Ci = − (kθ + | Pi−Pe |) (5) In the equation (5), θ is an angle (referred to as a displacement angle) between the predicted position vector Pe and the candidate position vector Pi, as shown in FIG. It is). The adjustment coefficient (arbitrary coefficient) k may be a scalar quantity and may be a constant, but may be changed in consideration of the stability of the past velocity vector direction of the storage object. For example, when the direction of the velocity vector is stable, the value of the adjustment coefficient k is increased to strengthen the influence of the displacement angle θ, and conversely, when the direction of the velocity is unstable,
By reducing the value of the adjustment coefficient k to weaken the influence of the displacement angle, it is possible to weaken the so-called directivity of the correspondence degree Ci.

In the equation (5), the displacement angle θ can be obtained by the equation (6), and the displacement | Pi-Pe | can be obtained by the equation (7).

Θ = arcCOS (Pe · Pi / | Pe || Pi |) = arcCOS {PexPix + PeyPii / √ (Pex ² + Pey ² ) √ (Pix ² + Pyy ² )} (6) | Pi-Pe | = √ { (Pix-Pex) ² + (Piy-Pey) ² } (7) The displacement angle θ is not limited to the inner product of the candidate position vector Pi and the predicted position vector Pe, but the inner product of the candidate position vector Pi and the velocity vector V. It goes without saying that it can also be obtained by the inverse calculation of.

As can be readily understood from the equations (6) and (7), the equation (5) is much more troublesome than the equation (1). It can be seen that the calculation cost of the degree of correspondence Ci according to (5) is much smaller than the calculation cost of the degree of correspondence Ci according to (5).

Further, the present invention is not limited to the above-mentioned embodiments, and it goes without saying that various configurations can be adopted without departing from the gist of the present invention.

[0087]

As described above, according to the present invention, a detected object that is closest to the predicted position of an object whose velocity and position are known after a lapse of a fixed time is simply stored in the object (memory Object), and evaluate the degree of correspondence of the detected object by taking into account the displacement angle from the velocity vector direction based on the position of the memory object and the distance between the predicted position and the detected position. Therefore, the effect that the object can be tracked more accurately is achieved.

In this case, the degree of correspondence is, for example, a predicted position vector having the known position of the object as the starting point and the predicted position as the ending point, and a plurality of candidate position vectors having the known position of the object as the starting point and the detected position as the ending point. The calculation is easy if the value obtained by subtracting the distance between the predicted position and the detected position (or the distance obtained by multiplying this distance by a constant) from the inner product value of.

Further, even if an object is lost due to an interruption or the like at a certain detection time point, if the record of the storage object is saved, at the next detection time point or the next detection time point, The effect of being able to detect a lost object again is achieved.

More specifically, the ID of the preceding vehicle or the obstacle is detected even if there is an intersection of the trails of the preceding vehicle or the obstacle or noise.
The effect that is correctly inherited without being accidentally inherited is also achieved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of functional blocks according to an embodiment of the present invention.

FIG. 2 is a diagram provided for explaining the content of a record.

FIG. 3 is a diagram provided for explaining a connection structure using a record pointer.

FIG. 4 is a diagram provided for explaining a predicted position vector and a search range.

FIG. 5 is a diagram provided for explaining calculation of a degree of correspondence between a predicted position vector and a candidate position vector.

FIG. 6 is a diagram showing the correspondence of detected objects.

FIG. 7 is a diagram provided for explaining an operation of erroneous detection of an object.

FIG. 8 is a block diagram showing a specific configuration of an embodiment of the present invention.

FIG. 9 is an overall flowchart of a tracking processing routine used for explaining the operation of the embodiment of the present invention.

FIG. 10 is a detailed flowchart of correspondence determination processing.

FIG. 11 is a detailed flowchart of a storage update process.

FIG. 12 is a diagram provided for explaining a case where an incorrect succession of ID occurs when left and right preceding vehicles are replaced, and A is
FIG. 3B is a diagram showing a state before the replacement, and FIG. 3B is a diagram showing a state in which the ID is incorrectly inherited after the replacement.

FIG. 13 is a diagram provided for explaining a case where an incorrect succession of an ID occurs when a front vehicle in front and a rear vehicle are replaced with each other.
FIG. 3B is a diagram showing a state before the replacement, and FIG. 3B is a diagram showing a state in which the ID is incorrectly inherited after the replacement.

FIG. 14 is a diagram which is used for explaining a case where an erroneous succession of an ID occurs when a front vehicle in the front side is aligned with a front vehicle in the front side among front and rear front vehicles. FIG. 4A is a diagram showing a state before being aligned, and FIG. B is a diagram showing a state in which an ID has been erroneously inherited after being aligned.

[Explanation of symbols]

8 ... Object detection device 10 ... Position detection means 11 ... Storage means 12 ... Correspondence degree calculation means 13 ... Correspondence determination means 14 ... Velocity calculation means 15 ... Memory update means 20 ... Position detection device 21 ... Computer 30 ... Moving obstacle detection device 51, 51E, 51P ... Record 53P, 53P ', 53P ", 54P ... Storage object 54, 55, 56 ... Detected object 58 ... Search range PP ... Reference past position PE ... Predicted position Pe ... Predicted position vector Pi ... Candidate position vector θ: displacement angle

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G05D 1/02 G05D 1/02 J G08G 1/00 J G08G 1/00 1/16 E 1/16 H04N 7/18 G // H04N 7/18 G01S 13/93 Z

Claims (5)

[Claims] 1. An object position detecting means for detecting a current position of a relatively moving object at regular time intervals, and a speed of the detected object based on the detected current position and past position of the object and the constant time. A velocity vector calculating means for calculating a vector; a storage means for storing the current position and velocity vector of the detected object as a pair; and a current position and velocity vector of the stored detected object (hereinafter referred to as a storage object). Based on the current position as a reference past position, a prediction means for calculating a prediction position vector having the reference past position as a starting point and the predicted arrival position of the storage object after the elapse of the certain time as an end point, and the object position A position vector calculated by calculating the position vector with the current position of the object newly detected by the detection means as the end point and the reference past position of the storage object as the start point, Among these, the object detection device is provided with a correspondence determination unit that identifies a position vector corresponding to the predicted position vector as a position vector representing the current position of the storage object. 2. An object position detecting means for detecting a current position of a relatively moving object at regular time intervals, and a speed of the detected object based on the detected current position and past position of the object and the constant time. A velocity vector calculation unit that calculates a vector, a storage unit that stores the current position and velocity vector of the detected object as a pair, and a current position and velocity of the stored detected object (hereinafter referred to as a stored object). And a prediction unit that calculates a predicted position vector having the current past position as a reference past position, the reference past position as a start point, and the predicted arrival position of the storage object after the elapse of the certain time as an end point, based on the vector; A position vector having the current position of the object newly detected by the object position detection means as the end point and the reference past position of the storage object as the start point is calculated, and calculated. Among the position vectors, the correspondence determining means that identifies the position vector corresponding to the predicted position vector as the position vector representing the current position of the storage object, and the current position represented by the tip of the identified position vector, An object detection apparatus comprising: a storage update unit that updates the content of the record of the storage object with a pair of the identified position vector and a velocity vector calculated based on the fixed time. 3. The storage updating means, in the correspondence determining means, deletes the record of the storage object for which the corresponding identified position vector does not exist a predetermined number of times from the storage means, and identifies the corresponding position. 3. The object detection device according to claim 2, wherein the current position and velocity vector of the newly detected object having no position vector are stored as a new record in the storage means as a new record. 4. The correspondence determining means determines a predicted position vector Pe, and a position vector having the current position of the object newly detected by the object position detecting means as an end point and the reference past position of the memory object as a start point. Pi, an arbitrary coefficient k, and a correspondence degree Ci, the correspondence degree Ci is Ci = Pe · Pi
4. The object according to claim 2, wherein the position vector Pi is represented by −k | Pi−Pe |, and the position vector Pi having the maximum value of the correspondence Ci is identified as a position vector corresponding to the predicted position vector Pe. Detection device. 5. An object detection method used for detecting and tracking a relatively moving object, wherein a predicted position of the object after a certain period of time is calculated from a velocity obtained at a known position of the object. Predicted position calculation process, at a time point after the elapse of the certain time, an object position detection process of detecting the position of the object, a direction connecting the known position of the object and the predicted position of the object, and the known position of the object And a correspondence angle calculation process of calculating a displacement angle formed by a direction connecting the detection position of the object, a distance between the predicted position of the object and the detection position of the object, and the displacement angle is small, And a determination step of determining that the position of the detected object having the small distance is the actual position of the object whose known position is known after a certain period of time elapses.