JP5655497B2 - Obstacle recognition device and obstacle recognition method - Google Patents

Description

  The present invention relates to an obstacle recognition apparatus that performs obstacle recognition using measurement results of a plurality of sensors.

  Conventionally, a technique called sensor fusion is known in which the measurement results of a plurality of sensors having different detection principles are integrated to recognize an obstacle. For example, Patent Document 1 discloses a preceding vehicle recognition apparatus and method that recognizes a preceding vehicle by sensor fusion based on a distance measurement result of a distance measurement sensor and a captured image of an image sensor.

JP-A-2005-90974

  Here, in the preceding vehicle recognition apparatus as described above, a radio wave radar is generally used as a distance measuring sensor. However, since the radio wave radar has a low horizontal resolution, it can separate a plurality of obstacles. In some cases, it may be mistakenly recognized as a single obstacle. In that case, there is a problem that consistency between the detection results of the radio wave radar and other sensors cannot be obtained, and appropriate obstacle recognition cannot be performed.

  Therefore, an object of the present invention is to provide an obstacle recognition device that can recognize an obstacle with high accuracy by matching the obstacle detection results of a plurality of sensors.

In order to solve the above-described problem, an obstacle recognition apparatus according to the present invention includes a first obstacle detection unit that transmits an electromagnetic wave and detects an obstacle based on a reflected output of the reflected wave, and a first obstacle. A second obstacle detection means for detecting an obstacle by a method different from the detection means, and a synthetic estimation in which a plurality of obstacles detected by the second obstacle detection means are combined with estimated reflection outputs of the obstacles. Reflection output estimation means for calculating a reflection output, and obstacle integrated recognition means for recognizing an obstacle using a comparison result between the combined estimated reflection output and the measured reflection output measured by the first obstacle detection means, The obstacle integrated recognizing means includes a plurality of obstacles that cannot be detected separately by the first obstacle detecting means when the difference between the combined estimated reflected output and the measured reflected output is equal to or less than a predetermined value. In the detection result by the object detection means Recognizing separated Zui.

  In this way, by calculating the combined estimated reflection output based on the obstacle detection result by the second obstacle detection means and recognizing the obstacle using the comparison result between the combined estimated reflection output and the measured reflection output, Even when there are obstacles that cannot be detected separately by the first obstacle detection means, the obstacle detection results of the first obstacle detection means and the obstacle detection results of the second obstacle detection means are inconsistent. It becomes possible to perform appropriate obstacle recognition. Thereby, an obstacle can be recognized with high accuracy.

  Moreover, it is preferable to further include an obstacle grouping processing means for grouping a plurality of obstacles that are predicted to be undetectable by the first obstacle detection means.

  By providing such an obstacle grouping processing means, obstacles that can be detected separately can be excluded from the detection target of the estimated reflection output, so that the processing load in obstacle recognition can be reduced. It becomes possible.

  In performing obstacle recognition in the obstacle recognition device, if the difference between the combined estimated reflection output and the measured reflection output is equal to or less than a predetermined value, the obstacle detection result by the second obstacle detection means is appropriate. Therefore, the obstacle is separated and recognized based on the detection result by the second obstacle detection means. Thereby, even when a plurality of adjacent obstacles cannot be separated by the first obstacle detection means and are detected as a single obstacle, the obstacles are appropriately recognized by the second obstacle detection means. be able to. In addition, since the estimated reflection output and the measured reflection output are matched, the obstacle can be recognized with higher accuracy than the obstacle detection only by the second obstacle detection means. As described above, according to the obstacle recognition apparatus according to the present invention, the first obstacle detection unit and the second obstacle detection unit suppress the inconsistency between the obstacle detection results, and recognize the obstacle with high accuracy. can do.

The reflected output estimating means has a reflected output model of the reflected wave from the obstacle, and predicts the estimated reflected output based on the width of the obstacle detected by the second obstacle detecting means and the reflected output model. It is preferable to do.

  When the width of the obstacle is wide, the reflected output of the reflected wave from the obstacle is large, and when the width of the obstacle is narrow, the reflected output of the reflected wave from the obstacle is small. Therefore, the estimated reflected output can be appropriately predicted by using the width of the obstacle and the reflected output model corresponding to the obstacle.

Further, the reflected output estimating means has a reflected output model of a reflected wave from the obstacle, and the estimated reflection is based on the kind and width of the obstacle detected by the second obstacle detecting means and the reflected output model. It is preferable to predict the output.

  The reflected output of the reflected wave with respect to the transmitted electromagnetic wave differs depending on the type of obstacle such as a person and a car. Therefore, in addition to the obstacle width and reflection output model, the estimated reflection output can be appropriately predicted by using the type of the obstacle.

  The reflected output model is preferably a normal distribution function.

  It is known that the reflected output of the reflected wave with respect to the obstacle has a normal distribution based on the width of the obstacle. Therefore, the estimated reflection output can be appropriately calculated by using a normal distribution function as the reflection output model.

  The obstacle recognition apparatus according to the present invention is preferably mounted on a vehicle and recognizes obstacles around the vehicle.

  By mounting the obstacle recognition apparatus according to the present invention on a vehicle and recognizing obstacles around the vehicle with high accuracy, collision between the host vehicle and obstacles existing around the vehicle can be avoided appropriately, and safety is ensured. Can increase the sex.

Also, the obstacle recognition method according to the present invention is different from the first obstacle detection process and the first obstacle detection process in which an electromagnetic wave is transmitted and an obstacle is detected based on the reflected output of the reflected wave. A second obstacle detection step of detecting an obstacle by the method, and a combined estimated reflection output obtained by synthesizing the estimated reflection outputs of the obstacles for the plurality of obstacles detected in the second obstacle detection step. a reflective output estimating step, seen containing an obstacle integrated recognition step of recognizing an obstacle by using the comparison result between the measured reflection output that is measured in the synthesis estimated reflection output and the first obstacle detection process, the obstacle In the integrated recognition step, when the difference between the combined estimated reflection output and the measured reflection output is equal to or less than a predetermined value, the second obstacle detection step detects a plurality of obstacles that cannot be separated and detected in the first obstacle detection step. In the detection result in Recognizing separated Zui.

  Thus, in the obstacle recognition method according to the present invention, the combined estimated reflection output is calculated based on the obstacle detection result in the second obstacle detection step, and the comparison result between the combined estimated reflection output and the measured reflection output is calculated. The obstacle detection result in the first obstacle detection step and the second obstacle detection step even when there is an obstacle that cannot be separated in the first obstacle detection step by recognizing the obstacle using It is possible to suppress the inconsistency of the obstacle detection results in the above and to perform appropriate obstacle detection. Thereby, an obstacle can be recognized with high accuracy.

  ADVANTAGE OF THE INVENTION According to this invention, the obstacle recognition apparatus which can recognize an obstacle with high precision can be provided by matching the obstacle detection result by a some sensor.

It is a block block diagram of the obstruction recognition apparatus which concerns on embodiment of this invention. It is a flowchart about operation | movement of the obstruction recognition apparatus which concerns on embodiment of this invention. It is a figure which shows the pairing process of the last predicted value and this detection result. It is a figure which shows the example which calculates the non-separable area | region estimated that a millimeter wave radar cannot be separated. It is a figure which shows the example which groups a several obstruction. FIG. 6A is a diagram showing an example of the estimated reflection output when the obstacle is a car, and FIG. 6B is a diagram showing an example of the estimated reflection output when the obstacle is a person. It is a graph which shows the relationship between the width | variety of an obstruction, and the peak value of an estimated reflected output. It is a figure which shows the synthetic | combination estimation reflection output of the obstruction grouped.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

  FIG. 1 is a schematic configuration diagram of an obstacle recognition apparatus according to an embodiment of the present invention.

  The obstacle recognition apparatus 1 according to the present embodiment is an apparatus that is mounted on a vehicle and recognizes obstacles existing around the vehicle using measurement results of a plurality of sensors. The parking assist device 1 includes a millimeter wave radar 11 (first obstacle detection means), a camera 12 (second obstacle detection means), and an obstacle recognition ECU (Electronic Control Unit) 2.

  A millimeter wave radar 11 and a camera 12 are connected to the obstacle recognition ECU 2. The millimeter wave radar 11 radiates radio waves in front of the vehicle, receives radio waves reflected from the surface of the object, detects the presence / absence of an object from the frequency change of the received signal, the direction of the object viewed from the vehicle, and the distance from the vehicle to the object Information such as the size of the object and the relative speed of the object with respect to the vehicle is acquired. Further, the millimeter wave radar 11 sends various information related to the detected object to the obstacle recognition ECU 2.

  The camera 12 images the front of the vehicle and acquires a captured image of the front of the vehicle. In addition, the camera 12 sends the acquired captured image in front of the vehicle to the obstacle recognition ECU 2.

  The obstacle recognition ECU 2 includes, for example, a CPU, a ROM, a RAM, and the like, and includes an obstacle detection unit 21, an obstacle grouping processing unit 22 (obstacle grouping processing unit), and an estimated reflection output calculation unit 23 (reflection output estimation unit). And an integrated recognition processing unit 24 (obstacle integrated recognition means). The obstacle detection unit 21 detects a region including an obstacle around the host vehicle based on the radar information output from the millimeter wave radar 11. Further, based on the image information output from the camera 12, a region including an obstacle is detected and the type of the obstacle is recognized. Obstacle information detected by the camera 12 is output to the obstacle grouping processing unit 22.

  The obstacle grouping processing unit 22 may detect a single obstacle from the obstacle detection result based on the image information, because the obstacle detection based on the radar information cannot separate a plurality of obstacles. Detect obstacles and group them. The grouped obstacle information is output to the estimated reflection output calculation unit 23.

  The estimated reflection output calculation unit 23 predicts the estimated reflection output from information such as the type and width of the obstacles grouped. Further, a combined estimated reflection output obtained by combining the estimated reflection outputs predicted for each obstacle is calculated, and the result is output to the integrated recognition processing unit 24.

  The integrated recognition processing unit 24 recognizes an obstacle existing around the host vehicle using both the obstacle detection result based on the millimeter wave radar 11 and the obstacle detection result based on the camera 12. Further, the combined estimated reflected output and the reflected output measured by the millimeter wave radar 11 are compared, and when the difference is equal to or smaller than a predetermined value, the obstacle is separated and recognized using the obstacle detection result by the camera 12. Here, the predetermined value is a set value set in the ECU 2 in advance. This value is determined depending on the detection characteristics of the millimeter wave radar 11 and the camera 12.

  The obstacle recognition apparatus 1 is preferably used as a part of a vehicle driving support system for avoiding a vehicle collision. In this vehicle driving support system, based on the obstacle detection information detected by the obstacle recognition device 1, a warning process for the driver of the vehicle, a brake preparation process for avoiding a collision, and a braking control for avoiding a collision are performed. Execute part or all. By executing these processes, it is possible to avoid collisions with obstacles or reduce the impact of collisions with obstacles.

  Next, the operation of the obstacle recognition apparatus 1 according to this embodiment will be described.

  FIG. 2 is a flowchart of the operation of the obstacle recognition apparatus according to the embodiment of the present invention. The control process in FIG. 2 is repeatedly executed at a predetermined period set in advance by the obstacle recognition ECU 2, for example.

  First, as shown in FIG. 2, pairing of the previous predicted value and the current detection result is performed (S10). The pairing of the previous predicted value and the current detection result is a process of associating the previously detected obstacle with the obstacle detected this time by the obstacle detection unit 21. A pairing method of the previous predicted value and the current detection result will be described with reference to FIG. The previous predicted value T in FIG. 3 is the predicted position of the obstacle at the current time predicted from the relative speed between the obstacle found in the previous process and the vehicle, position information, and the like. The detection results Rn and Cn (n is a positive integer) are detection positions of obstacles detected by reading the measurement results of the millimeter wave radar 11 and the camera 12, respectively. When the current detection result R1 by the millimeter wave radar 11 and the current detection result C1 by the camera 12 are within the allowable error range A with respect to the previous predicted value T, the obstacle detected in the previous process The detection results R1 and C1 this time are determined to be the same obstacle.

  In addition, when one of the current detection result R1 by the millimeter wave radar 11 and the current detection result C1 by the camera 12 is not detected within the allowable error range A, the consistency of the detection results cannot be obtained among a plurality of sensors. For this reason, it is determined that the reliability of the detection result is low, and the obstacle recognition is not performed.

  Various methods are known for setting the allowable error range A. In this embodiment, an ellipse εc in which the error in the front-rear direction predicted in the obstacle detection by the camera is the major axis and the error in the left-right direction is the minor axis. And an ellipse εr having a minor axis as an error in the front-rear direction and a major axis as an error in the left-right direction, which are predicted in obstacle detection by the millimeter wave radar 11, and an ellipse having the major axis of the ellipse εc and εr as an allowable error are set. Range A is set.

  Next, a new obstacle process is performed from the previous predicted value acquired in S10 and the current detection result (S11). If the previous predicted values cannot be paired as in the current detection results R2 and C2 shown in FIG. 3, the current detection results R2 and C2 are set and stored as new obstacle candidates.

  Subsequently, the obstacle grouping processing unit 22 performs an obstacle grouping process (S12). In the obstacle grouping process, the millimeter wave radar may not be able to separate multiple obstacles from the paired obstacles and new obstacle candidates, and may be erroneously recognized as a single obstacle. Obstacle candidates are grouped. The obstacle grouping process is performed using the obstacle detection result obtained by the camera 12.

  FIG. 4 is an example of calculating the non-separable region F predicted that the millimeter wave radar cannot be separated. As shown in FIG. 4, the sum of the width W of the obstacle detected from the camera 12, the left and right portions of the detection error σw in the width direction of the camera 12, and the left and right portions of the millimeter wave non-separable distance d is separated. The width of the impossible area F is set. Further, the height of the obstacle detected by the camera 12 plus the vertical portion of the detection error σh in the height direction of the camera 12 is set as the height of the non-separable area F. Here, the detection errors σw and σh in the width direction of the camera 12 and the millimeter wave non-separable distance d may be values determined depending on the detection characteristics of the camera and millimeter wave radar to be used, respectively.

  Next, the non-separable area F is set in a plurality of obstacles having the same distance from the host vehicle in the paired obstacle and the new obstacle candidate. Here, the same range includes not only the same distance from the host vehicle to the obstacle but also the substantially same distance that the millimeter wave radar 11 cannot separate in the distance direction. If there is an overlapping area between the non-separable area F and other obstacles, grouping is performed. FIG. 5 shows an example of performing obstacle grouping processing for a plurality of obstacles. Since the non-separable area F1 of the obstacle B1 belonging to the obstacle group G overlaps with a part of the other obstacle B2, the obstacle B2 is added to the obstacle group G. Furthermore, since the non-separable area F2 of the obstacle B2 belonging to the obstacle group G overlaps with a part of the other obstacle B3, the obstacle B3 is added to the obstacle group G. Since the non-separable area F3 of the obstacle B3 belonging to the obstacle group G overlaps with a part of the other obstacle B4, the obstacle B4 is added to the obstacle group G.

  Subsequently, an estimated reflection output calculation process is performed on the obstacle group G grouped in S12 (S13). In the estimated reflection output calculation process, the estimated reflection output of each grouped obstacle is calculated, and the estimated reflection outputs of all the grouped obstacles are added together to calculate the combined estimated reflection output. The FIG. 6 shows an example of calculating the estimated reflection output of each obstacle. The estimated reflection output is calculated as a normal distribution function with the standard deviation as the width of the obstacle detected by the camera 12. The peak value of the estimated reflection output is set based on the type of obstacle (for example, a car or a person). FIG. 6A is an example of the estimated reflection output when the obstacle is a car, and FIG. 6B is an example of the estimated reflection output when the obstacle is a person. As shown in FIGS. 6A and 6B, the peak value Pcar of the estimated reflected output when the obstacle is a car is compared with the peak value Pped of the estimated reflected output when the obstacle is a person. It is set large. Even when the obstacle is a car, it is preferable to vary the peak value Pcar based on the width of the car. A table showing the relationship between the type and width of the obstacle and the peak value of the estimated reflected output is stored in advance in the ROM of the obstacle recognition ECU 2, and the obstacle recognition ECU 2 reads the table from the ROM as necessary and estimates it. The reflected output is calculated.

  Further, when a sensor other than a camera such as a laser radar is used as the sensor and the type of obstacle cannot be recognized, the peak value of the estimated reflected output may be set according to the width of the obstacle. . FIG. 7 is a graph showing the relationship between the width of an obstacle and the peak value of the corresponding estimated reflected output.

  After the estimated reflection outputs of the individual obstacles of the grouped obstacle group G are calculated, the estimated reflection outputs of all the obstacles of the obstacle group G are added as shown in FIG. Reflected output. Further, a point at which the differential value becomes 0 from the calculated estimated reflection output, that is, a peak value is set as a millimeter wave estimated output candidate. In the example of FIG. 8, P1 and P2 are millimeter wave estimation output candidates.

  When the combined estimated reflection output is calculated in S13, the process proceeds to recognition result integration processing (S14). In the recognition result integration process, first, the calculated combined estimated reflection output and the reflection output measured by the millimeter wave radar 11 are compared. Specifically, a plurality of peak values in the measured reflected output are compared with the millimeter wave estimated output candidates, and when the difference is equal to or smaller than a predetermined value, the estimated reflected output and the measured reflected output are matched. To be judged. Here, the predetermined value is a value determined depending on the detection characteristics of the millimeter wave radar 11 and the camera 12.

  When the estimated reflection output and the measured reflection output are matched, the obstacles are recognized by separating the obstacles using the obstacle detection result by the camera 12 for the obstacle groups grouped by the obstacle grouping process. The After that, pairing of the detected obstacle and the previous predicted value T is performed, whereby obstacle tracking and new obstacle detection are performed.

  Next, the process proceeds to the next value prediction process (S15). In the next value prediction process, the predicted position of the obstacle in the next obstacle recognition process is estimated from the obstacle information recognized in S14 using a Kalman filter or the like. As a method for estimating the predicted position of the obstacle after such a predetermined time has been proposed, a detailed method is omitted in this specification. The estimated predicted position of the obstacle is used for pairing in the obstacle recognition process after a predetermined period.

  As described above, according to the obstacle recognition apparatus 1 according to the present embodiment, the combined estimated reflection output is calculated based on the obstacle detection result by the camera 12 and compared with the measurement reflection output by the millimeter wave radar 11. When there is an obstacle that cannot be detected separately by the millimeter wave radar 11, the obstacle is separated and recognized using the obstacle detection result of the camera 12. As a result, the obstacle detection result of the camera 12 and the obstacle detection result of the millimeter wave radar 11 are prevented from being inconsistent, and appropriate obstacle detection can be performed. Thereby, an obstacle can be recognized with high accuracy.

  In addition, although preferred embodiment of this invention was described, this invention is not limited to the said embodiment. For example, in the present embodiment, obstacle recognition is performed using the millimeter wave radar 11 and the camera 12, but obstacle detection is performed using a sensor having excellent lateral resolution such as a laser radar instead of the camera 12. It is also possible to do this.

  In the present embodiment, in the obstacle grouping process, the grouping is performed when there is an overlapping area between the non-separable area F and other obstacles. However, the non-separable area F and the non-separable area between other obstacles are not included. Obstacles may be grouped when F overlaps.

  Furthermore, in the present embodiment, matching between the combined estimated reflected output and the measured reflected output is performed by the difference between the plurality of peak values in the measured reflected output and the millimeter wave estimated output candidates, but different methods are used. May be. For example, an integrated value of the combined estimated reflected output and the measured reflected output is obtained in a predetermined range, and matching can be determined based on whether or not the difference between them is equal to or less than a predetermined value.

DESCRIPTION OF SYMBOLS 1 ... Obstacle recognition apparatus, 2 ... Obstacle recognition ECU, 11 ... Millimeter wave radar, 12 ... Camera, 21 ... Obstacle detection part, 22 ... Obstacle grouping process part, 23 ... Estimated reflection output calculation part, 24 ... Integration Recognition processing unit.

Claims (7)

First obstacle detection means for transmitting an electromagnetic wave and detecting an obstacle based on the reflected output of the reflected wave;
Second obstacle detection means for detecting an obstacle by a method different from the first obstacle detection means;
Reflection output estimation means for calculating a combined estimated reflection output obtained by combining the estimated reflection outputs of the obstacles detected by the second obstacle detection means;
An obstacle integrated recognition unit that recognizes an obstacle using a comparison result between the combined estimated reflection output and the measured reflection output measured by the first obstacle detection unit ;
When the difference between the combined estimated reflected output and the measured reflected output is less than or equal to a predetermined value, the obstacle integrated recognition means detects a plurality of obstacles that cannot be separated and detected by the first obstacle detecting means. Recognizing separately based on the detection result of 2 obstacle detection means,
Obstacle recognition device.   The obstacle recognition apparatus according to claim 1, further comprising obstacle grouping processing means for grouping a plurality of obstacles predicted not to be detected separately by the first obstacle detection means. The reflected output estimating means has a reflected output model of a reflected wave from an obstacle, and the estimated reflected output is calculated based on the width of the obstacle detected by the second obstacle detecting means and the reflected output model. It predicted to, obstacle recognition device according to claim 1 or 2. The reflected output estimating means has a reflected output model of a reflected wave from an obstacle, and the estimation is performed based on the type and width of the obstacle detected by the second obstacle detecting means and the reflected output model. The obstacle recognition apparatus according to claim 1 or 2 , which predicts a reflected output. The obstacle recognition apparatus according to claim 3 or 4 , wherein the reflected output model is a normal distribution function. The obstacle recognition apparatus according to any one of claims 1 to 5 , wherein the obstacle recognition apparatus is mounted on a vehicle and recognizes obstacles around the vehicle. A first obstacle detection step of transmitting an electromagnetic wave and detecting an obstacle based on the reflected output of the reflected wave;
A second obstacle detection step of detecting an obstacle by a method different from the first obstacle detection step;
A reflected output estimating step of calculating a combined estimated reflected output obtained by combining the estimated reflected outputs of the obstacles detected in the second obstacle detecting step;
See containing and a obstacle integrated recognition step of recognizing an obstacle by using the comparison result between the measured reflection output that is measured in the said synthesized estimate reflected output first obstacle detection process,
In the obstacle integrated recognition step, when a difference between the combined estimated reflection output and the measurement reflection output is a predetermined value or less, a plurality of obstacles that cannot be detected separately in the first obstacle detection step Recognizing separately based on the detection result in the second obstacle detection step,
Obstacle recognition method.

Images