JP5170058B2 - Object detection device - Google Patents

Description

  The present invention relates to an object detection device.

  A technique is known in which a different reference pattern is stored for each environment around the host vehicle, and a reference pattern corresponding to the environment around the host vehicle is selected when pattern matching is performed (see, for example, Patent Document 1). ).

  However, since a large number of surrounding environments are assumed and the reference pattern must be stored, a huge memory area is required. Moreover, if a reference pattern corresponding to any surrounding environment is prepared and pattern matching is performed, there is a possibility that false detection of an object may increase.

JP 2007-329762 A JP 2008-015770 A JP 2006-350670 A

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a technique for improving the accuracy of pattern matching while suppressing the amount of data necessary for performing pattern matching.

In order to achieve the above object, the object detection apparatus according to the present invention employs the following means. That is, the object detection device according to the present invention is
Distance detection means for detecting the distance from the vehicle to the object;
Irradiating means for irradiating light around the host vehicle;
Imaging means for imaging the object;
A range detection means for detecting a light irradiation range by the irradiation means;
Storage means for storing a reference shape of an object divided into a plurality of regions;
Weighting means for weighting the reference shape of the object for each of the plurality of regions based on the distance detected by the distance detecting means and the irradiation range detected by the range detecting means;
Matching means for performing object matching by comparing the data of the object imaged by the imaging means and the reference shape of the object weighted by the weighting means for each of the plurality of regions;
It is characterized by providing.

  The matching means performs, for example, pattern matching between the object data to be imaged and the weighted reference shape of each area. Here, when the host vehicle is irradiating light when traveling at night, traveling in a tunnel, or the like, the illuminated area changes in the object by changing the irradiation range of the light. If it does so, the data of the object imaged will change according to the irradiation range of light. The irradiation range can be a range in which the illuminance is a specified value or more.

When a part of the object is located within the light irradiation range, the shape of the part of the object is easily recognized, but the shape of the other part of the object located outside the irradiation range is difficult to recognize. Then, since the reliability of the data of the object belonging to the region that is difficult to recognize becomes low, there is a high possibility that the object is mistaken when using the data of this region. For example, even if it is determined that the degree of similarity between the data of the object imaged by the imaging unit and the reference shape is high outside the light irradiation range, there is a possibility that the degree of similarity is determined to be high due to misperception. is there. Therefore, the weighting unit weights the reference shape for each of the plurality of regions in accordance with the light irradiation range. For example, the weight is increased as the position where the shape of the object is easily recognized. Thereby, the influence degree of the reference | standard shape of the position which is easy to recognize becomes larger. Then, since the influence degree of the reference shape at a position that is difficult to recognize becomes small, it is possible to prevent the object from being misidentified.

  In this way, even in a region where the degree of similarity is high, the degree of influence on the whole is small in a region where the weight is small. Since the object can be identified according to the degree of similarity and the weight by weighting the reference shape based on the irradiation range, the object identification accuracy can be improved.

  In the present invention, the distance detection unit and the imaging unit may include a stereo camera.

  According to the stereo camera, an object can be imaged. Further, according to the two data obtained by the stereo camera, the distance to the object can be calculated. That is, if a stereo camera is provided, there is no need to provide a radar or the like for measuring the distance to the object.

  In this invention, the said irradiation means may be comprised including the headlamp of the said own vehicle. That is, since the light irradiation range changes according to the state of the headlamp, the range detection unit can easily determine the irradiation range according to the light irradiation range by the irradiation unit. The irradiating means may include an auxiliary lamp that illuminates the front or side of the host vehicle, or a back lamp that illuminates the rear of the host vehicle.

  In the present invention, the range detection means stores an irradiation range according to the direction of the headlamp, and can select the irradiation range according to the direction of the headlamp.

  Here, if the direction of the headlamp changes, the light irradiation range also changes. Thereby, the area | region which enters into the irradiation range on an object may also change. If the light irradiation range corresponding to the direction of the headlamp is stored in advance, the light irradiation range can be easily obtained by detecting the direction of the headlamp.

The headlamp changes the irradiation range in the vertical direction,
The range detecting means stores an irradiation range when the headlamp is upward and an irradiation range when the headlamp is downward, and selects the irradiation range according to whether the headlamp is upward or downward. Can do.

  That is, the light irradiation range changes depending on whether the headlamp is upward or downward. Thereby, the area | region which enters into the irradiation range on an object may also change. If the irradiation range of light in the case where the headlamp is upward and downward is stored in advance, the light irradiation range can be easily obtained by detecting whether the headlamp is upward or downward. .

In addition, the headlamp changes the irradiation range in the left-right direction,
The range detection means stores each irradiation range when the headlamp changes in the left-right direction, and can select the irradiation range according to which direction the headlamp is in the left-right direction. it can.

  For example, the headlight direction may change in the left-right direction while the vehicle is traveling on a curve. In such a case, the light irradiation range also changes as the direction of the headlamp changes in the left-right direction. Thereby, the area | region which enters into the irradiation range on an object may also change. If the light irradiation range corresponding to the direction of the headlight in the left-right direction is stored in advance, the light irradiation range can be simplified by detecting which direction the headlight is in the left-right direction. Can get to.

  In the present invention, the weighting means can increase the weight in a region with higher illuminance. That is, the higher the illuminance, the easier it is to recognize the shape of the object, so the reliability of the imaged object data is higher, so the weight is increased.

  ADVANTAGE OF THE INVENTION According to this invention, the precision of pattern matching can be improved, suppressing the data amount required in order to perform pattern matching.

It is a block diagram which shows the object detection apparatus which concerns on an Example. It is the figure which showed the irradiation range of the light by a headlamp. It is the figure which showed the weight of division | segmentation data. It is the flowchart which showed the flow of the matching process which concerns on an Example. It is a figure which shows an example of the pedestrian irradiation range table for LO, or the pedestrian irradiation range table for HI.

  Hereinafter, specific embodiments of the object detection apparatus according to the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram illustrating an object detection apparatus 1 according to the present embodiment. The object detection apparatus 1 according to the present embodiment is an apparatus that is mounted on a host vehicle traveling on a road and detects objects such as pedestrians and other vehicles existing around the host vehicle.

  The object detection device 1 includes a stereo camera 2, a headlamp switch 3, a headlamp 4, and an ECU 5.

  The stereo camera 2 is provided in the front part or the room of the host vehicle and images an object existing in front of the host vehicle. The stereo camera 2 can record distance information to the object by simultaneously imaging the object from a plurality of different directions. The stereo camera 2 converts light coming from the lens into an electrical signal and sequentially outputs it to the ECU 5. In this embodiment, the stereo camera 2 corresponds to the image pickup means in the present invention.

  The headlamp switch 3 is a switch for the driver to select ON (lit) -OFF (dark) and HI (upward) -LO (downward) of the headlamp 4. Note that the headlamp 4 may be automatically turned on and off according to the brightness around the host vehicle. A signal corresponding to the state of the headlamp switch 3 is input to the ECU 5, and the headlamp 4 is controlled according to the state.

  The headlamp 4 is a lamp that can be switched at least between upward (HI) and downward (LO). In this embodiment, the headlamp 4 corresponds to the irradiation means in the present invention.

  The ECU 5 determines whether or not an object is present in front of the host vehicle or the type of the object, and is configured mainly by a computer including a CPU, a ROM, and a RAM, for example. The ECU 5 includes a divided data storage unit 51, a distance calculation unit 52, a light irradiation range calculation unit 53, a weighting calculation unit 54, and a matching calculation unit 55.

  The divided data storage unit 51 stores the divided data obtained by dividing the basic shape of the object into a plurality (for example, 6). The basic shape is data to be compared with the data of the object imaged by the stereo camera 2. The divided data storage unit 51 stores the reference shape of the object divided into a plurality of areas for each area. In the present embodiment, the divided data storage unit 51 corresponds to the storage means in the present invention.

  The distance calculation unit 52 calculates the distance between the host vehicle and the object based on the data of the object imaged by the stereo camera 2. Note that when the object is positioned obliquely forward of the host vehicle, the actual distance is divided into a traveling direction component of the host vehicle and a true lateral component (direction component orthogonal to the traveling direction) of the host vehicle. Then, the component in the traveling direction of the host vehicle is used as the distance, and the component in the right lateral direction on the right side of the host vehicle is determined as the lateral position. The height of the object is also obtained at the same time. When the object is located on the left front side of the host vehicle, the lateral position is a negative value.

  The light irradiation range calculation unit 53 calculates the irradiation range of the headlamp 4 based on the signal from the headlamp switch 3. Since the light irradiation range changes depending on the state of the headlamp 4 (HI and LO states of the headlamp switch 3), the relationship between the state of the headlamp switch 3 or the state of the headlamp 4 and the light irradiation range is Each is obtained by experiment or the like and stored in the ECU 5. In the present embodiment, the light irradiation range calculation unit 53 corresponds to the range detection means in the present invention.

  The weighting calculation unit 54 weights the divided data stored in the divided data storage unit 51 based on the calculation result of the light irradiation range calculation unit 53. The relationship between the light irradiation range and the data weight is obtained in advance by experiments or the like. In this embodiment, the weighting calculation unit 54 corresponds to the weighting means in the present invention.

  The matching calculation unit 55 compares the divided data after being weighted by the weighting calculation unit 54 with the data of the object imaged by the stereo camera 2, and stores the reference shape of the stored object and the imaging The degree of similarity with the data of the selected object is calculated. In this embodiment, the matching calculation unit 55 corresponds to the matching means in the present invention.

  Then, when the matching degree is equal to or greater than the specified value, the ECU 5 outputs a signal to the operation device 6. The operation device 6 is, for example, an LED 61 or an alarm device 62 using sound. The actuating device 6 can also include controlling the brake and engine.

  In the present embodiment, a pedestrian will be described as an example as an object, but the present invention can be similarly applied to other objects.

  Here, when the vehicle is traveling at night or the like and the surroundings of the host vehicle are dark, the appearance of the pedestrian changes depending on the state of the headlamp 4, that is, whether it is upward or downward. That is, if the state of the headlamp 4 is different, even if pedestrians existing at the same position are imaged, the data is not the same. Then, when pattern matching is performed using the reference shape stored in the divided data storage unit 51 as it is, the identification result of the pedestrian can change depending on the state of the headlamp 4.

  FIG. 2 is a diagram showing a light irradiation range by the headlamp 4. The irradiation range is, for example, a range where the illuminance is a specified value or more. FIG. 2A shows a case where the headlamp 4 is downward, and FIG. 2B shows a case where the headlamp 4 is upward. Pedestrians are represented as X, Y, and Z in order of increasing distance from the host vehicle 100. 2A and 2B differ only in the state of the headlamp 4, and the distance between the host vehicle 100 and the pedestrians X, Y, and Z is the same. For the pedestrian X, the lower body enters the irradiation range when the headlamp 4 is downward, and the upper body enters the irradiation range when the headlamp 4 is upward. Further, when the headlamp 4 is downward, the pedestrian Y enters the irradiation range at the lower part of the foot, and when the headlamp 4 is upward, the whole body enters the irradiation range. The pedestrian Z does not enter the light irradiation range even if the headlamp 4 is downward or upward.

  Thus, the part which enters into the irradiation range of the light from the headlamp 4 differs depending on the state of the headlamp 4, the distance from the own vehicle 100 to the pedestrian, and the lateral position. And since it is hard to recognize a pedestrian outside the light irradiation range, the reliability of the data imaged with the stereo camera 2 is low. For this reason, if pattern matching is performed using data outside the light irradiation range as it is, the pedestrian identification accuracy may be lowered.

  Therefore, in the present embodiment, the divided data is weighted, and the weight of the position where the light from the headlamp 4 hits more (the position where the illuminance is larger) is increased. Since this weight varies depending on the light irradiation range, it varies depending on the position, size, and luminous intensity of the headlamp 4, and is obtained in advance through experiments or the like.

  FIG. 3 is a diagram showing the weight of the divided data. FIG. 3A shows a case where the headlamp 4 is downward, and FIG. 3B shows a case where the headlamp 4 is upward. FIG. 3 shows a case where the distance from the host vehicle to the pedestrian (traveling direction component) is 15 m and the lateral position (right direction component) is −1 m. That is, the pedestrian is located diagonally to the left of the host vehicle.

  In FIG. 3, the reference shape of the pedestrian is divided into six, and the weight is set in each region. In addition, since the position where the light from the headlamp 4 strikes and the position where it does not strike are not clearly separated, weighting is performed according to, for example, illuminance. In the example of FIG. 3, when the headlamp 4 is facing downward (FIG. 3A), light is mainly applied to the lower body of the pedestrian. Get smaller. The weight is set to 0 in the area including the head. Here, the weight of 0 means that the data in this area among the captured pedestrian data does not affect the pedestrian identification.

  On the other hand, when the headlamp 4 is facing upward (FIG. 3B), the entire pedestrian is exposed to light, so the weight of the entire region increases.

  In this way, the weighting calculation unit 54 weights the divided data according to the state of the headlamp 4, the distance from the host vehicle to the pedestrian, and the lateral position.

  The matching calculation unit 55 divides the pedestrian data captured by the stereo camera 2 into six parts, and performs pattern matching between the captured pedestrian data and the weighted divided data in each region. Identify pedestrians with For example, a score obtained by multiplying the similarity and the weight in each area is calculated, and whether or not the person is a pedestrian is identified by the sum of the scores in all the areas.

  FIG. 4 is a flowchart showing a flow of matching processing according to the present embodiment. This routine is repeatedly executed every predetermined time.

In step S101, the distance calculation unit 52 acquires the distance, the lateral position, and the height from the host vehicle to the center of the pedestrian based on the data captured by the stereo camera 2.

If the three-dimensional information of the distance, the lateral position, and the height can be obtained, the coordinates (CX, CY) of the two-dimensional image can be obtained.
(CX, CY) = FuncComvert3Dto2D (distance, lateral position, pedestrian height)
However, FuncComvert3Dto2D is a function that converts three-dimensional information into coordinates of a two-dimensional image.

  The left lateral position and the right lateral position of the pedestrian are calculated by adding a specified offset amount to the lateral direction of the lateral position thus obtained. This offset amount is set so that a predetermined value is added to the width of the pedestrian between the left lateral position and the right lateral position, and is, for example, ± 0.6 m. That is, the left lateral position and the right lateral position are set so that the entire pedestrian enters between the left lateral position and the right lateral position. Similarly, a predetermined offset amount is added in the vertical direction from the center of the pedestrian to calculate the pedestrian's height position and ground position. This offset amount is a value obtained by adding a predetermined value to the height of the center of the pedestrian, and is, for example, ± 1.2 m. That is, the ground position and the height position of the pedestrian are set so that the entire pedestrian enters between the ground position and the height position of the pedestrian.

That is, the upper left point (LX, UY), lower left point (LX, DY), upper right point (RX, UY), lower right point (RX, DY) when a pedestrian on a two-dimensional image is surrounded by a rectangle is as follows: It is obtained like this.
(LX, UY) = FuncComvert3Dto2D (distance, left lateral position, pedestrian height)
(LX, DY) = FuncComvert3Dto2D (distance, left lateral position, ground)
(RX, UY) = FuncComvert3Dto2D (distance, right lateral position, pedestrian height)
(RX, DY) = FuncComvert3Dto2D (distance, right lateral position, ground)
The area surrounded by the four points is set as a matching area, and is compared with the divided data stored in the divided data storage unit 51.

  Next, in step S102, the state of the headlamp 4 is acquired. That is, it is acquired whether the headlamp 4 is turned on, turned off, or turned on when the headlamp 4 is turned on.

  Here, when the headlamp 4 is turned off, the divided data is not weighted. That is, the divided data stored in the divided data storage unit 51 is used as it is. When the headlamp 4 is downward, the “LO pedestrian irradiation range table” is referred to. When the headlamp 4 is upward, the “HI pedestrian irradiation range table” is referred to, and the light irradiation range of the headlamp 4 is acquired. .

  FIG. 5 is a diagram illustrating an example of the LO pedestrian irradiation range table or the HI pedestrian irradiation range table. In the LO pedestrian irradiation range table and the HI pedestrian irradiation range table, the height of the pedestrian's head and foot is set in accordance with the distance from the vehicle to the pedestrian. The height of the foot is the height from the ground to the lower limit of the irradiation range, and the height of the head is the height from the ground to the upper limit of the irradiation range. In the case of FIG. 5, when the distance from the own vehicle to the pedestrian is 15 m, the irradiation range is from the foot height (0.1 m) to the head height (1.0 m).

In step S103, the divided data is weighted. In this embodiment, the reference shape is divided into six regions vertically. Each divided area is weighted based on the height of the head and the height of the foot. The weighting is a value from 0 to 10, for example. Here, by calculating (head height / pedestrian height) and (foot height / pedestrian height), the head or foot set in FIG. 5 belongs to any divided region. Can ask.

  For example, the weighting is performed as follows. (1) The weight of an area that does not correspond to the position of the head or foot and is within the range of the position of the foot from the position of the head is set to 10. (2) The area other than (1) is set to 0 in an area not corresponding to the position of the head or foot. (3) In the area corresponding to the head or foot, the inside of the area is further subdivided into 9 parts, and from the position of the head or foot to the area defined in (1) and (2) above in the subdivided area A value from 1 to 9 is used while changing the weight. In addition, since weighting should just be performed according to an irradiation range and illumination intensity, you may perform it by another method.

  In step S104, pedestrians are identified by matching processing. Pattern matching is performed between the weighted divided data and the data in the matching area to identify pedestrians. A score obtained by multiplying the similarity and the weight for each divided region of the reference shape is calculated, and whether or not the person is a pedestrian is identified by a total score obtained by adding the scores for each region. Note that the similarity can be obtained by a known method. Since the probability of being a pedestrian increases as the total score increases, for example, when the total score is greater than or equal to a threshold value, the person is identified as a pedestrian. And when it is identified as a pedestrian, the operation device 6 is operated.

  As described above, according to the present embodiment, since the reference shape is weighted based on the light irradiation range of the headlamp 4, it is not necessary to store a large amount of data in advance. That is, the memory capacity can be reduced. In addition, by performing weighting based on the light irradiation range, the weight of the region where the object is easily recognized increases, so that the object identification accuracy can be improved.

  In the present embodiment, an example in which the headlamp 4 that illuminates the front of the host vehicle has been described, but the present invention can be similarly applied to a backlamp that illuminates the rear of the host vehicle. Moreover, it can apply similarly about the auxiliary lamp or corner lamp which illuminates the side or the front of the own vehicle.

  Moreover, although the example which changes the headlamp 4 in two steps of upward and downward was demonstrated, even if it changes in three steps or more, and also changes steplessly, it can apply similarly. In this case, the light irradiation range may be stored in accordance with each stage. Moreover, although the example which changes the headlamp 4 by the up-down direction of upward and downward was demonstrated, it can apply similarly, even if it changes to the left-right direction. For example, the present invention can be similarly applied to a variable light distribution type headlamp system that directs the optical axis in the steering direction during cornering. Also in this case, the irradiation range of light is stored according to each direction.

  In this embodiment, the example in which the distance from the own vehicle to the object is detected based on the data captured by the stereo camera 2 has been described. However, the distance to the object may be detected by a radar.

  Moreover, although the present Example demonstrated the example which calculates the irradiation range of light according to the state of the headlight 4, you may obtain the irradiation range of light by detecting the illumination intensity of a road surface with a sensor.

DESCRIPTION OF SYMBOLS 1 Object detection apparatus 2 Stereo camera 3 Headlamp switch 4 Headlamp 5 ECU
6 Actuating Device 51 Division Data Storage Unit 52 Distance Calculation Unit 53 Light Irradiation Range Calculation Unit 54 Weighting Calculation Unit 55 Matching Calculation Unit 61 LED
62 Alarm device 100 Own vehicle

Claims (7)

Distance detection means for detecting the distance from the vehicle to the object;
Irradiating means for irradiating light around the host vehicle;
Imaging means for imaging the object;
A range detection means for detecting a light irradiation range by the irradiation means;
Storage means for storing a reference shape of an object divided into a plurality of regions;
Weighting means for weighting the reference shape of the object for each of the plurality of regions based on the distance detected by the distance detecting means and the irradiation range detected by the range detecting means;
Matching means for performing object matching by comparing the data of the object imaged by the imaging means and the reference shape of the object weighted by the weighting means for each of the plurality of regions;
An object detection apparatus comprising:   The object detection apparatus according to claim 1, wherein the distance detection unit and the imaging unit include a stereo camera.   The object detection apparatus according to claim 1, wherein the irradiation unit includes a headlamp of the host vehicle.   The object detection according to claim 3, wherein the range detection unit stores an irradiation range corresponding to the direction of the headlamp, and selects the irradiation range according to the direction of the headlamp. apparatus. The headlamp changes the irradiation range in the vertical direction,
The range detecting means stores an irradiation range when the headlamp is upward and an irradiation range when the headlamp is downward, and selects the irradiation range according to whether the headlamp is upward or downward. The object detection apparatus according to claim 4. The headlamp changes the irradiation range in the left-right direction,
The range detection means stores each irradiation range when the headlamp changes in the left-right direction, and selects the irradiation range according to which direction the headlamp is in the left-right direction. The object detection device according to claim 4, wherein   7. The object detection apparatus according to claim 1, wherein the weighting unit increases the weight in a region with higher illuminance.

Images