JP2006047033A - Object recognition method and device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To recognize an object in front of a self vehicle while adding the determination of vehicle likeness from the results of searching ahead of the vehicle while mounting only a scanning laser radar. <P>SOLUTION: An object recognition device searches the forward of the self vehicle repeatedly by the scanning laser radar 2 mounted on the self vehicle 1, detects the observation value of the vehicle index regarding the position of the body in front of the self vehicle from which the laser radar 2 receives reflection light, calculates the recognition evaluation value reversely varying to the difference between the predicted value of the vehicle index estimated from the observation and observed value detected this time, and recognizes the body by calculating the probability on the basis of the proportional characteristics set up on the recognition evaluation value and the probability being a vehicle. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an object recognition method and an object recognition apparatus for recognizing an object in front of a host vehicle from a search result in front of the host vehicle by a scanning laser radar mounted on the host vehicle.

  Conventionally, in a vehicle equipped with a vehicle driving support system called ACC (Adaptive Cruise Control), a pre-crash safety system, etc., an approach to an obstacle such as a preceding vehicle ahead of the host vehicle is warned, or so-called damage reduction automatic braking is performed. In order to realize functions, etc., an imaging device represented by a scanning laser radar or a monocular camera is mounted on the host vehicle as an obstacle detection sensor in front of the host vehicle, and an object in front of the host vehicle is recognized. (For example, refer to Patent Document 1).

  When a scanning laser radar is mounted on the host vehicle as the detection sensor, the front of the host vehicle is searched by scanning the front of the host vehicle in the vehicle width direction (left-right direction) with laser pulse irradiation. By repeating the above, the presence / absence of an object as an obstacle is detected based on the presence / absence of reception of a reflected wave by a reflector or the like of a vehicle (a so-called four-wheel vehicle or two-wheel vehicle) to recognize an object in front of the host vehicle.

  In this case, the scanning laser radar is used as a distance measuring sensor for the distance to the object, and when the reflected wave is received, the distance from the vehicle to the point of the reflected wave is determined based on the transmission / reception time difference of the laser pulse. Calculated and measured.

  On the other hand, when only the imaging device is mounted on the host vehicle as the detection sensor, the front of the host vehicle is continuously shot by the imaging device, so-called optical flow is detected by image processing such as edge detection of the shot image, and the It has been proposed to recognize an obstacle such as a preceding vehicle in front of the vehicle separately from a non-obstacle consisting of road structures such as roadside signs, road signs, guardrails, and the like (see, for example, Patent Document 2).

  In addition, a scanning laser radar and an imaging device are mounted on the vehicle to form a so-called sensor fusion vehicle front detection sensor, and information on the front of the vehicle is obtained by scanning radar imaging and imaging. It has also been proposed to recognize obstacles ahead of the host vehicle such as a preceding vehicle with the possibility of being distinguished from the non-obstacles more accurately (see, for example, Patent Document 3).

Japanese Unexamined Patent Publication No. 2000-344033 (paragraphs [0012] and [0016], FIG. 3) JP 11-353565 (paragraphs [0022]-[0025], [0048], FIG. 1) JP 7-182484 A (paragraphs [0006]-[0010], FIG. 1)

  In the case of the conventional configuration in which the scanning laser radar is mounted on the vehicle as the vehicle front detection sensor, when the reflected wave is received, it is unconditionally recognized that the vehicle as an obstacle exists in front of the vehicle. Even when reflected waves from non-obstacles such as signboards, road signs and guardrails are received, there is a problem that these non-obstacles are mistakenly recognized as obstacle vehicles. Therefore, there is a risk of false alarms and over alarms for approaching obstacles, and erroneous control and over control of automatic braking.

  On the other hand, in the case of the conventional configuration in which the conventional imaging device is mounted on the own vehicle alone or in combination with the scanning laser radar, the obstacle and the non-obstacle can be distinguished and recognized from the captured image, An expensive imaging device is required, and complicated image processing is also required, resulting in a problem of an expensive and complicated configuration.

  In this type of object recognition, from the viewpoint of improving safety with an inexpensive configuration as much as possible, only a scanning laser radar is mounted on the vehicle as a vehicle front detection sensor, It is desirable to recognize an object.

  The present invention mounts only a scanning laser radar on the own vehicle, and recognizes an object in front of the own vehicle from the search result in front of the own vehicle, taking into account the determination of the likelihood of the vehicle. The purpose is to prevent as much as possible misrecognition.

  In order to achieve the above-described object, the object recognition method of the present invention repeatedly searches the front of the vehicle with a scanning laser radar mounted on the vehicle, and the laser radar receives reflected light based on the search. Detect an observed value of the vehicle index related to the position of the object ahead of the host vehicle, and increase or decrease the difference from the difference between the predicted value of the vehicle index estimated from the already detected observation value and the observed value detected this time On the contrary, the recognition evaluation value that changes in level is calculated, the probability of the object is calculated based on a set proportional characteristic of the recognition evaluation value and the probability of being a vehicle, and the object is recognized. (Claim 1).

  In this case, the observed value of the vehicle index related to the position of the object ahead of the host vehicle is the distance from the host vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar, and the predicted value of the vehicle index is It is preferable to calculate and estimate from the change characteristics of the distance (Claim 2), and the observed value of the vehicle index related to the position of the object in front of the own vehicle is calculated from the transmission / reception time difference of the scanning laser radar. It is the relative speed of the object based on the time change of the distance from the vehicle to the object, and it is preferable to estimate the predicted value of the vehicle index by calculating from the change characteristic of the relative speed.

  Next, in the object recognition method of the present invention, an object in front of the vehicle in which the front of the vehicle is repeatedly searched by a scanning laser radar mounted on the vehicle, and the laser radar receives reflected light based on the search. Detecting an observation value of a vehicle index related to the size of the vehicle, and calculating a recognition evaluation value that changes in elevation in reverse to the increase or decrease of the difference from the difference between the observation value detected this time and the observation value detected last time, The probability of the object is calculated based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and the object is recognized (claim 4).

  In this case, it is preferable that the observation value of the vehicle index related to the size of the object ahead of the host vehicle is the lateral width of the object detected from the interval of the reception position of the reflected light of the scanning laser radar. .

  Further, the object recognition method of the present invention repeatedly searches the front of the vehicle by a scanning laser radar mounted on the vehicle, and detects an observation value of a vehicle index related to the reception state of the laser radar based on the search. Then, a recognition evaluation value that changes in height according to the difference between the observed value detected this time and the set reference value is calculated, and based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, The object is recognized by calculating the probability of the object in front of the vehicle (claim 6).

  In this case, the observation value of the vehicle index related to the reception state of the scanning laser radar of the reflected light by the object in front of the host vehicle is the reflected light reception intensity of the scanning laser radar, and the set reference value is It is preferable that it is an average value of the reflected light receiving intensity when the vehicle is used (claim 7).

  Furthermore, the object recognition method of the present invention repeatedly searches the front of the vehicle by a scanning laser radar mounted on the vehicle, and based on the search, the position of the object in front of the vehicle where the laser radar receives reflected light. , Dimensions, and observation values of a plurality of vehicle indicators related to the respective reception states of the reflected light of the laser radar are detected, and the observation values detected this time, estimated prediction values, previous detection values, set criteria Each recognition evaluation value that changes in height according to a difference from each comparison value consisting of each value is calculated, and based on a set proportional characteristic between the average value of each recognition evaluation value and the probability of being a vehicle, The object is recognized by calculating the probability (claim 8).

  In this case, the observed value of the vehicle index related to the position of the object in front of the vehicle is the distance from the own vehicle to the object in front of the own vehicle calculated from the transmission / reception time difference of the scanning laser radar, and the time variation of the distance Each of the relative speeds of the object, and the observed value of the vehicle index related to the size of the object is the lateral width of the object detected from the interval of the light receiving position of the reflected light of the scanning laser radar, and the reception of the reflected light The observation value of the vehicle index related to the state is the reflected light reception intensity of the scanning laser radar, and the comparison value of the distance and the relative velocity is the distance based on the already detected observation value and the relative velocity, respectively. It consists of predicted values of the vehicle index for each of the distance and the relative speed estimated from the change characteristics, and the comparison value of the lateral width is the observation value of the lateral width detected last time, and the reflected light reception Comparison of the intensity is, the object is preferably made of a reference value the set to the average value of the reflected light reception intensity when the vehicle (claim 9).

  Further, the recognition evaluation value of each observation value of the vehicle index is normalized so that the variation of each recognition evaluation value is substantially equal, and the average value of the respective recognition evaluation values normalized and the probability of being a vehicle are set. It is desirable to calculate the probability of an object ahead of the host vehicle based on the proportional characteristic.

  Next, an object recognition apparatus according to the present invention includes a scanning laser radar that is mounted on a host vehicle and repeatedly searches in front of the host vehicle, and an object in front of the host vehicle that receives reflected light based on the search. An index detection unit for detecting an observed value of the vehicle index related to the position; a predicted value of the vehicle index is estimated from the observation value already detected by the index detection unit; and the predicted value and the index detection unit detect this time A recognition evaluation value calculating means for calculating a recognition evaluation value that changes in height in the opposite direction of the increase / decrease of the difference, and a set proportional characteristic of the recognition evaluation value and the probability of being a vehicle. And a recognition output means for calculating the probability of the object and outputting the calculation result as a recognition result of the object (claim 11).

  In this case, the index detection means detects the distance from the vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar as an observation value of the vehicle index related to the position of the object ahead of the vehicle, and recognizes and evaluates it. Preferably, the predicted value of the vehicle index is calculated and estimated from the change characteristic of the distance by the value calculation means (Claim 12), and the vehicle related to the position of the object ahead of the host vehicle by the index detection means. As the observed value of the index, the relative speed of the object is detected based on the time change of the distance from the own vehicle to the object, which is calculated from the transmission / reception time difference of the scanning laser radar. It is also preferable to calculate and estimate the predicted value from the change characteristic of the relative speed (claim 13).

  Next, an object recognition apparatus according to the present invention includes a scanning laser radar that is mounted on a host vehicle and repeatedly searches in front of the host vehicle, and an object in front of the host vehicle that receives reflected light based on the search. An index detection means for detecting an observation value of a vehicle index related to the size, and a difference between the observation value detected this time and the observation value detected last time of the index detection means, the level changes in reverse to the increase or decrease of the difference. A recognition evaluation value calculating means for calculating a recognition evaluation value; and calculating the probability of the object based on a set proportional characteristic of the recognition evaluation value and the probability of being a vehicle, and using the calculation result as the recognition result of the object And a recognition output means for outputting (claim 14).

  In this case, it is preferable that the width of the object is detected by the index detection means from the interval between the positions where the reflected light of the scanning laser radar is received as the observation value of the vehicle index related to the size of the object ahead of the host vehicle. 15).

  In addition, the object recognition device of the present invention detects a measured value of a vehicle index related to a reception state of the laser radar based on the scanning and a scanning laser radar mounted on the own vehicle to repeatedly search in front of the own vehicle. Index recognition means, recognition recognition value calculation means for calculating a recognition evaluation value that changes depending on a difference between the observed value detected this time by the index detection means and a set reference value, and the recognition evaluation value It comprises a recognition output means for calculating the probability of an object ahead of the host vehicle based on a set proportional characteristic with the probability of being a vehicle and outputting the calculation result as a recognition result of the object ( Claim 16).

  In this case, the index detection means detects the reflected light reception intensity of the scanning laser radar as an observed value of the vehicle index related to the reception state of the reflected laser scanning laser radar, and the recognition evaluation value calculation means automatically detects the reflected light reception intensity. It is preferable to set an average value of the reflected light reception intensity when the object in front of the vehicle is a vehicle.

  Furthermore, the object recognition device of the present invention is equipped with a scanning laser radar that is mounted on the host vehicle and repeatedly searches the front of the host vehicle, and an object in front of the host vehicle that receives the reflected light based on the search. Estimating the position, size, and indicator detection means for detecting observation values of a plurality of vehicle indicators related to the reflected light reception states of the laser radar, and the observation values detected by the indicator detection means at this time A recognition evaluation value calculating means for calculating each recognition evaluation value that changes depending on a difference from each comparison value including a predicted value, a previous detection value, and a set reference value; and an average value of each recognition evaluation value Recognizing output means for calculating the probability of the object based on a set proportional characteristic with the probability of being a vehicle and outputting the calculation result as a recognition result of the object is provided. .

  In this case, the distance from the own vehicle to the object ahead of the host vehicle calculated from the transmission / reception time difference of the scanning laser radar as the observed value of the vehicle index related to the position of the object ahead of the host vehicle by the index detection means, the distance The relative speed of the object is detected based on the time variation of the object, and the lateral width of the object detected from the interval of the reception position of the reflected light of the laser radar is detected as an observation value of the vehicle index related to the size of the object Then, the reflected light received intensity of the laser radar is detected as an observation value of the vehicle index related to the reception state of the reflected light of the laser radar, and the comparison value of the distance and the relative speed is detected by the recognition evaluation value calculation means. The predicted value of the vehicle index for each of the distance and the relative speed is calculated from the change characteristics of the distance and the relative speed based on the already detected observation value, and the comparison value of the lateral width is calculated. Forming a comparison value of the reflected light received intensity of the laser radar as a reference value set to an average value of the reflected light received intensity when the object is a vehicle. Preferred (claim 19).

  Further, the recognition evaluation value calculation means normalizes the recognition evaluation value of each observation value of the vehicle index so that variations of the respective recognition evaluation values are substantially equal, and the recognition output means normalizes each of the recognition evaluation values. It is desirable to calculate the probability of an object ahead of the vehicle based on a set proportional characteristic between the average value of the vehicle and the probability of being a vehicle.

  First, according to the configuration of claims 1 and 11, the observed value of the vehicle index related to the position of the object in front of the vehicle is based on the repeated search in front of the vehicle by the scanning laser radar mounted on the vehicle. If the object ahead of the host vehicle is a vehicle (preceding vehicle) as an obstacle at this time, the observed value is detected based on the reflected light at substantially the same location of the vehicle, and the position of the vehicle Since it changes continuously according to the change, the predicted value of the vehicle index detected in the next search can be estimated stably and accurately from the change characteristic of the observed value.

  On the other hand, if the object in front of the host vehicle is a road structure such as a roadside signboard or guardrail that is a non-obstacle, the object will be instantaneously removed from the search range due to the vehicle traveling, Since the position of the reflected light changes from time to time and reception of the reflected light becomes irregular, the predicted value detected in the next search cannot be estimated from the change characteristic of the observed value.

  Therefore, the recognition evaluation value calculated from the difference between the predicted value of the vehicle index estimated from the previously detected observation value and the currently detected observation value is high when the object ahead of the host vehicle is a vehicle as an obstacle. It becomes low at the time of the said non-obstacle.

  And, since the probability of vehicle likelihood (vehicle probability) increases in proportion to the recognition evaluation value, the vehicle probability calculated from the recognition evaluation value increases when the object ahead of the vehicle is a vehicle as an obstacle, Low for road structures such as roadside signs and guardrails on non-obstacles.

  Therefore, based on the vehicle probability calculated based on the search result of the scanning laser radar in front of the vehicle, it is possible to recognize the object in front of the vehicle by taking into account the determination of the vehicle, and based on this recognition, the roadside It is possible to prevent as many as possible non-obstacles such as road signs such as signs and guardrails from being mistakenly recognized as obstacle vehicles, and only a low-priced, small scanning laser radar can be used as a detection sensor in front of the vehicle. With the configuration, the recognition performance of an object in front of the host vehicle can be improved to improve the alarm and control of the ACC and the pre-crash safety system, thereby improving safety.

  Next, according to the second and twelfth aspects of the present invention, the observed value of the vehicle index related to the position of the object ahead of the host vehicle is calculated from the own vehicle calculated from the transmission / reception time difference of the scanning laser radar. By using the distance to the object, when the object ahead of the host vehicle is a vehicle (preceding vehicle), the predicted value of the vehicle index can be calculated and estimated from the change characteristics of the distance with extremely high accuracy. The accuracy of the vehicle probability is improved.

  According to the third and thirteenth aspects of the present invention, the observed value of the vehicle index is calculated based on the relative speed of the object ahead of the host vehicle based on the time change of the distance calculated from the transmission / reception time difference of the scanning laser radar. Thus, when the object ahead of the host vehicle is a vehicle (preceding vehicle), the predicted value of the vehicle index can be calculated and estimated with extremely high accuracy as in the case of the distance. As in the case of the 12 configuration, the accuracy of the vehicle probability is improved.

  Next, according to the configurations of claims 4 and 14, the observed value of the vehicle index related to the size of the object in front of the own vehicle based on the repeated search in front of the own vehicle of the scanning laser radar mounted on the own vehicle. In this case, if the object ahead of the host vehicle is a vehicle (preceding vehicle) as an obstacle, the observation value detected this time and the observation value detected last time are almost the same because the vehicle does not change constantly. It will be the same and the difference will be smaller.

  On the other hand, if the object in front of the vehicle is a road structure such as a roadside signboard or guardrail that is a non-obstacle, the position of the reflected light changes from time to time as the vehicle travels. The detected value is different and the difference is large.

  Therefore, the recognition evaluation value calculated from the difference between the observation value detected this time and the observation value detected last time becomes high when the object ahead of the vehicle is a vehicle as an obstacle, and the non-obstacle The vehicle probability calculated from the recognition evaluation value is high when the object in front of the vehicle is a vehicle as an obstacle, and is a road structure such as a signboard or guardrail on the roadside of a non-obstacle. It becomes low.

  Therefore, as in the case of the observation value related to the position, the object in front of the vehicle is recognized based on the vehicle probability calculated based on the search result of the scanning laser radar in front of the vehicle, taking into consideration the likelihood of the vehicle. Based on this recognition, a situation in which a non-obstacle such as a roadside signboard or a guard structure such as a road structure is erroneously recognized as an obstacle vehicle can be prevented as much as possible. The same effect as the case can be obtained.

  Next, according to the configurations of claims 5 and 15, the observed value of the vehicle index related to the size of the object ahead of the host vehicle is detected from the interval between the light receiving positions of the reflected light of the scanning laser radar. By setting the width of the object, the vehicle probability can be obtained with extremely high accuracy based on practical observation values.

  Next, according to the configurations of claims 6 and 16, the scanning laser radar mounted on the own vehicle has a high brightness of the reflector if the object ahead of the own vehicle is a vehicle (preceding vehicle) as an obstacle. The reflected light is repeatedly received, and the observed value of the vehicle index related to the reception state is stable and large, but the object ahead of the vehicle is a non-obstacle object such as a roadside signboard or guardrail. If there is, the intensity and position of the reflected light changes from time to time, the reception of the reflected light becomes irregular, and the observed value of the vehicle index related to the reception state is small and fluctuates.

  Therefore, the difference between the observed value of the vehicle index related to the reception state detected this time and the set reference value becomes large (or small) when the object ahead of the vehicle is a vehicle as an obstacle, The recognition evaluation value is high when the object ahead of the host vehicle is a vehicle as an obstacle, and is low when the object is a non-obstacle.

  And, since the probability of vehicle likelihood (vehicle probability) increases in proportion to the recognition evaluation value, the vehicle probability calculated from the recognition evaluation value increases when the object ahead of the vehicle is a vehicle as an obstacle, Low for road structures such as roadside signs and guardrails on non-obstacles.

  Therefore, as in the case of the observed value of the vehicle index related to the position and the observed value of the vehicle index related to the size, the object in front of the host vehicle is recognized based on the vehicle probability in consideration of the vehicle-likeness determination. Based on this recognition, it is possible to prevent a situation in which a non-obstacle such as a roadside signboard or a guard structure such as a road structure is erroneously recognized as an obstacle vehicle as much as possible. The same effect as in the case of 14 can be obtained.

  Next, according to the configurations of claims 7 and 17, practical observation is performed by using the observation value of the vehicle index related to the reception state of the scanning laser radar as the reflected light reception intensity of the laser radar. The vehicle probability can be obtained with extremely high accuracy based on the value.

  Next, according to the configurations of claims 8 and 18, based on the repeated search in front of the own vehicle of the scanning laser radar mounted on the own vehicle, the position and size of the object in front of the own vehicle, the laser radar Observation values of a plurality of vehicle indices related to the respective reception states of the reflected light are detected, and each comparison value comprising each observation value detected this time, an estimated predicted value, a previous detection value, and a set reference value Each recognition evaluation value of the difference is calculated, and the vehicle probability of the object is calculated based on an average value of a plurality of recognition evaluation values with different observation targets.

  Therefore, the accuracy of the vehicle probability is remarkably improved as compared with the case based on one type of observation value, and the object ahead of the host vehicle can be recognized with higher accuracy in consideration of the vehicle-likeness determination. Based on this recognition, It is possible to reliably prevent a situation in which a non-obstacle such as a roadside signboard or a guardrail is mistakenly recognized as an obstacle vehicle, and an inexpensive, small scanning laser as a detection sensor in front of the vehicle. With the configuration in which only the radar is mounted, the recognition performance of the object ahead of the host vehicle can be remarkably improved, and the alarm and control of the ACC and the pre-crash safety system can be further improved, and the safety can be significantly improved.

  Next, according to the configurations of claims 9 and 19, an object in front of the host vehicle from the host vehicle in which the observation value of the vehicle index related to the position of the object in front of the host vehicle is calculated from the transmission / reception time difference of the scanning laser radar. The vehicle object observation value is detected from the interval of the received positions of the reflected light of the laser radar. The observation value of the vehicle index related to the reception state of the reflected light of the laser radar is composed of the reflected light reception intensity of the laser radar, and the distance, the comparison value of the speed is the distance, It consists of the predicted value of the vehicle index for each speed, the observed value of the lateral width detected last time, and the reference value of the average value of the reflected light received intensity when the object is a vehicle, so that it is extremely accurate with a practical configuration. Can be obtained Ku said vehicle probability.

  Next, according to the configuration of claims 10 and 20, the recognition evaluation value of each observation value of the vehicle index is normalized, and the normalized average value of each recognition evaluation value and the probability of being a vehicle are set. Since the probability of the object ahead of the vehicle is calculated based on the proportional characteristics, each recognition evaluation value is standardized to have the same variation even if the variation tendency of each observation value is different. The vehicle probability is calculated with the average recognition accuracy of each recognition evaluation value, and the vehicle probability can be obtained with higher accuracy to further improve the recognition performance of the object ahead of the host vehicle.

  Next, in order to describe the present invention in more detail, one embodiment thereof will be described in detail with reference to FIGS.

  FIG. 1 is a block diagram of an object recognition device for an own vehicle 1 having an approach warning / automatic brake control function such as an ACC and a pre-crash safety system, and FIGS. 2 and 3 are vehicle probability (vehicle probability) of FIG. FIG. 4 is a flowchart for explaining the operation of FIG. 1. FIG. 4 is a characteristic diagram of control timing based on the recognition result.

  The object recognition apparatus of FIG. 1 includes only an inexpensive and small-sized scanning laser radar 2 mounted on the own vehicle 1 as a detection sensor in front of the own vehicle. The laser radar 2 is provided at almost the vehicle center position such as the interior of the host vehicle 1 or in the hood, and repeatedly scans the front of the host vehicle in the vehicle width direction (left and right direction) while irradiating the laser with pulses. The front side of the vehicle is searched while receiving the reflected waves of the reflectors of the four-wheeled vehicle and the preceding vehicle of the two-wheeled vehicle.

  The object recognition apparatus also includes various sensors for detecting the vehicle state such as a vehicle speed sensor 3 having a wheel speed sensor configuration for detecting the vehicle speed.

  After the engine of the host vehicle 1 is started, the laser radar 2 repeatedly searches the front of the host vehicle and outputs a search result signal to the ECU 4 for object recognition / control processing of the microcomputer configuration.

  The ECU 4 executes a recognition process based on an object recognition program preset in the memory unit 5 and a timing control process for adjusting an ACC or pre-crash safety system alarm or an automatic brake intervention timing. The distance from the vehicle to the object ahead of the vehicle and the relative speed of the object are detected as the observation values related to the position, and the width of the object is detected as the observation value of the vehicle index related to the dimensions of the object. Then, the reflected light received intensity of the laser radar 2 is detected as an observation value of the vehicle index related to the reception state of the reflected light of the laser radar 2, and based on these detections, the probability of vehicle-likeness of the object ahead of the host vehicle ( Vehicle probability) is recognized, and the alarm / automatic brake control tie is detected by timing control processing based on the recognition result. To adjust the ring.

  Therefore, the ECU 4 forms the next index detection means 41, recognition evaluation value calculation means 42, recognition output means 43, and control timing adjustment means 44 by executing software processing.

(A) Index detection means 41
For each exploration of the laser radar 2, the means 41 determines the position and size of the object in front of the vehicle where the laser radar 2 receives the reflected light and the reception state of the reflected light of the laser radar 2 based on the result of the exploration. Detect observed values of multiple related vehicle indicators.

  That is, as an observation value of the vehicle index related to the position of the object in front of the host vehicle, the distance (position) from the host vehicle to the object in front of the host vehicle calculated from the transmission / reception time difference of the laser pulse of the laser radar 2, The relative speed of the object based on the time change is detected as an observation position (current position) Po and an observation speed (current speed) Vo, respectively, and the reflected light of the laser radar 2 is used as an observed value of the vehicle index related to the size of the object. The horizontal width of the object detected from the reception position interval is detected as the current detection size Sn, and the reflected light received intensity of the laser radar 2 is used as an observation value of the vehicle index related to the reception state of the reflected light of the laser radar 2. Detected as detected light reception intensity Pw.

(B) Recognition evaluation value calculation means 42
This means 42 changes in height according to the difference between each of the observed values Po to Pw detected this time by the index detection means 41 and each comparison value comprising the estimated value, the previous detection value, and the set reference value. Each recognition evaluation value is calculated.

  At this time, as the comparison values of the distance and the relative speed, the predicted values of the vehicle indices for the distance and the relative speed are obtained from the time variation characteristics of the distance and the relative speed, respectively, based on the already detected values Po and Vo. Calculated and estimated, and held in the memory unit 7 in a rewritable manner.

Specifically, when the scanning period of the laser radar 2 is T and the predicted value (predicted position) of the distance is Pe, the predicted value Pe is calculated as follows: predicted position Pe = previous observation value (previous position) + relative speed × It is calculated and estimated from the calculation of T + (relative acceleration × T ² ) / 2, and is stored in the memory unit 7 in a rewritable manner.

  If the relative speed (predicted speed) is Ve, the predicted speed Ve is calculated and estimated from the calculation of predicted speed Ve = previous speed + (relative acceleration × T), and is stored in the memory unit 7 in a rewritable manner. To do.

  Next, the comparison value of the lateral width of the object in front of the host vehicle is calculated, for example, by the difference in the coordinate value in the vehicle width direction (left-right direction) of the reflected light reception position that is equal to or higher than a certain reception level in the previous scan of the laser radar 2. Then, it is detected and held in the memory unit 7 in a rewritable manner as the previous detection size Sp.

  Further, the comparison value of the reflected light reception intensity of the laser radar 2 is the average value (distribution average value) PwSt of the reflected light of the reflector light of the vehicle located within a certain distance range ahead of the host vehicle obtained by various experiments. The average value PwSt is previously stored in the memory unit 7 as the reference value.

  Next, each recognition evaluation value corresponding to the difference between each observation value detected this time and each comparison value is specifically calculated as described below.

  First, assuming that the recognition evaluation values of the distance and the relative velocity are Pc and Vc, for the distance, the difference Perr between the observed value Po detected this time and the comparison value Pe is calculated as Perr = ABS (Pe−Po). (ABS is an operator of absolute value, the same applies hereinafter), and the recognition evaluation value Pc is obtained from the calculation of Pc = 1− (Perr × Kp) (Kp: adjustment gain).

  Similarly, for the relative velocity, the difference Verr between the observed value Vo detected this time and the comparison value Ve is obtained from the calculation of Verr = ABS (Ve−Vo), and the recognition evaluation value Vc is determined as Vc = 1− (Verr XKv) (Kv: adjustment gain).

  If the recognition evaluation value of the horizontal width is Sc, the difference Serr between the current detection size Sn and the previous detection size Sp is obtained from the calculation of Serr = ABS (Sp-Sn), and the recognition evaluation value Sc is Sc = 1. -Calculated from the calculation of (Serr × Ks) (Ks: adjustment gain).

  Further, when the recognition evaluation value of the reflected light reception intensity is Pwc, the difference Pwerr between the detected light reception intensity Pw and the reference value (average value PwSt) is obtained from the calculation of Pwerr = ABS (PwSt−Pw), and the recognition evaluation value. Pwc is obtained from the calculation of Pwc = 1− (Pwerr × Kpw) (Kpw: adjustment gain).

  In this case, the smaller the difference Perr to Pwerr is, the more the vehicle ahead of the vehicle is captured.

  Next, in this embodiment, the recognition evaluation values Pc to Pwc are normalized to 0 to 1 so that their variations are substantially equal to improve the recognition accuracy. The adjustment gains Kp to Kpw are tuned and set so as to converge to 0 ≦ Kp to Kpw ≦ 1.

  This tuning is performed, for example, by collecting various sample data of differences Perr to Pwerr through experiments or the like, and setting the adjustment gains Kp to Kpw so that the respective variances are equal from the distribution.

  Each of the recognition evaluation values Pc to Pwc increases when the object in front of the host vehicle is a vehicle as an obstacle, and the difference Perr to Pwerr decreases and increases. When the object is a non-obstacle, the difference Perr to Pwerr is It becomes larger and lowers, and changes positively opposite to the increase / decrease of the difference Perr to Pwerr.

(C) Recognition output means 43
This means 43 calculates the vehicle probability of the object ahead of the host vehicle based on the set proportional characteristic of the average value of each recognition evaluation value and the vehicle probability, and outputs the calculation result as the recognition result of the object.

  That is, the average value AveC (= (Pc + Po + Sc + Pwc) / 4) of the normalized recognition evaluation values Pc to Pwc of the recognition evaluation value calculation means 42 and the vehicle probability Car are determined based on, for example, various experiments. The lower limit value α of the average value AveC that sets the probability Car to 0 (assuming that it is not completely a vehicle) and the upper limit value β of the average value AveC that sets the vehicle probability Car to 1 (which is considered to be a complete vehicle) are determined (0 <α <Β <1), as a proportional characteristic in which the average value AveC and the vehicle probability Car are set, for example, data of the characteristic of the solid line A in FIG.

  The upper and lower limits α and β are determined from, for example, the distribution of a large number of histogram data of the average value AveC for vehicles and road structures.

  Each time the average value AveC is obtained by repeating the scanning of the laser radar 2, the corresponding vehicle probability Car is calculated from the characteristics of the solid line A, and the calculated vehicle probability Car is output as a recognition result.

(D) Control timing adjusting means 44
This means 44 is based on the vehicle probability Car of the recognition result of the recognition output means 43, for example, from the adjustment characteristic of the control timing of the solid line B in FIG. The intervention timing adjustment amount (delay amount) corresponding to the momentary vehicle probability Car is determined.

  In other words, the closer the vehicle probability of the recognition result is to 0, the higher the possibility that the object in front of the vehicle detected from the search result is a non-obstacle with a low possibility of collision. Control the alarm / automatic brake because there is a high possibility that the detected object in front of the vehicle is an obstacle vehicle as the vehicle probability of the recognition result is closer to 1. The timing adjustment amount is determined from the solid line b in FIG. 3 so that the timing is the original intervention timing or faster, and the control signal of the adjustment amount is sent to the alarm / automatic brake processing means (not shown) after the ECU 4. Output.

  When the above processing is shown in a flowchart, for example, as shown in steps S1 to S8 in FIG. 4, when the search output of the laser radar 2 is generated in step S1 while the vehicle 1 is traveling, the search output On the basis of the above, the index detection means 41 of the ECU 4 detects the observed values Po to Pw of the vehicle index related to the position and size of the object ahead of the host vehicle and the reception state of the reflected light of the laser radar 2, and the steps S2 to S5 Then, the recognition evaluation value calculation means 42 of the ECU 4 obtains the recognition evaluation values Pc to Pwc in the processing blocks 421 to 424 of the distance, speed, width, and received light intensity, and proceeds to step S6 to obtain the recognition evaluation values Pc to Pwc. Normalize to 0-1.

  In step S7, the recognition output means 43 of the ECU 4 calculates an average value AveC of the recognition evaluation values Pc to Pwc, calculates a corresponding vehicle probability Car, and uses the probability Car as a recognition result of an object ahead of the host vehicle. In step S8, based on the vehicle probability Car, the adjustment amount of the alarm / automatic brake control timing is determined according to the vehicle quality of the object ahead of the vehicle.

  Therefore, in the case of this embodiment, only an inexpensive and small laser radar 2 is mounted on the own vehicle 1 as a detection sensor in front of the own vehicle, the vehicle probability Car is obtained based on the search result, and the determination of the likelihood of the vehicle is taken into consideration. The object in front of the vehicle can be recognized, and based on the result of this recognition, the timing of the obstacle warning and automatic brake control is adjusted, and road signs such as road signs, road signs, guardrails, etc. Prevents false recognition of non-obstacles with no possibility of collisions as much as possible, and prevents false alarms and over-alarms of obstacle approach in ACC and pre-crash safety systems, and miscontrol and over-control of automatic brakes Thus, safety can be improved.

  Based on the observed values Po to Pw of a plurality of vehicle indices related to the position and size of the object in front of the vehicle and the reception state of the reflected light of the laser radar 2, a plurality of recognition evaluation values Pc to Since PWc is obtained and the vehicle probability Car is calculated based on the average value AveC, the accuracy of the vehicle probability is remarkably improved, and the object in front of the host vehicle can be recognized with a very high accuracy in consideration of the determination of the vehicle. On the basis of this recognition, it is possible to reliably prevent a situation in which a non-obstacle such as a roadside signboard or a guard structure such as a guard structure is erroneously recognized as an obstacle vehicle.

  The present invention is not limited to the above-described embodiment, and various modifications other than those described above can be made without departing from the spirit thereof, for example, the position of an object in front of the host vehicle, The observation values of the plurality of vehicle indices related to the dimensions and the reception state of the reflected light of the laser radar 2 are different from the distance, relative speed, lateral width, reflected light reception intensity of the embodiment, or observation values of those observation targets, or Of course, the observation values of a large number of observation targets obtained by adding new observation targets to each observation target of the embodiment may be used, and generally, the recognition accuracy improves as the number of observation targets increases.

  On the other hand, when simplification of processing, configuration, etc., the observation value of the vehicle index is, for example, any one or two of the position and size of the object in front of the host vehicle and the reception state of the reflected light of the laser radar 2, The number may be three, specifically, any one, two, or three of the distance, the relative speed, the lateral width, and the reflected light reception intensity.

  When the observed value of the vehicle index is, for example, one of the distance, relative speed, lateral width, and reflected light reception intensity, the calculation of the average value AveC and the normalization process are omitted, and the observation is performed. From the characteristic lines corresponding to the solid line A in FIG. 2 between the values Po,..., Pw and the vehicle probability Car, the vehicle probability Car is obtained and the object ahead of the host vehicle is recognized.

  Next, it is needless to say that the recognition result can be applied not only to the approach warning / automatic brake control described above but also to various travel controls of the vehicle 1.

  By the way, in order to reduce the number of equipment parts of the own vehicle 1, for example, the scanning laser radar 2 of FIG. 1 can be applied to other control sensors such as follow-up traveling control.

It is a block diagram of one embodiment of this invention. It is a characteristic view of an example of the probability (vehicle probability) of the vehicle likeness of FIG. It is explanatory drawing of an example of the adjustment characteristic of the timing control of FIG. It is a flowchart for process description of FIG.

Explanation of symbols

1 Vehicle 2 Scanning laser radar 4 ECU

Claims (20)

Repeated exploration in front of the vehicle with a scanning laser radar mounted on the vehicle,
Based on the exploration, the laser radar detects the observation value of the vehicle index related to the position of the object in front of the host vehicle that receives the reflected light,
From the difference between the predicted value of the vehicle index estimated from the detected value already detected and the observed value detected this time, calculate a recognition evaluation value that changes in level in the opposite direction of increase or decrease of the difference,
An object recognition method comprising: calculating the probability of the object based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and recognizing the object.   The observed value of the vehicle index related to the position of the object ahead of the host vehicle is the distance from the host vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar, and the predicted value of the vehicle index is the change in the distance. The object recognition method according to claim 1, wherein the object recognition method is calculated and estimated from characteristics.   The observed value of the vehicle index related to the position of the object ahead of the host vehicle is the relative speed of the object based on the time change of the distance from the host vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar, The object recognition method according to claim 1, wherein a predicted value of a vehicle index is estimated by calculating from a change characteristic of the relative speed. Repeated exploration in front of the vehicle with a scanning laser radar mounted on the vehicle,
Based on the exploration, the laser radar detects an observation value of the vehicle index related to the size of the object in front of the host vehicle that receives the reflected light,
From the difference between the observation value detected this time and the observation value detected last time, calculate a recognition evaluation value that changes in level opposite to the increase or decrease of the difference,
An object recognition method comprising: calculating the probability of the object based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and recognizing the object.   5. The object according to claim 4, wherein the observed value of the vehicle index related to the size of the object in front of the host vehicle is the width of the object detected from the interval of the reception position of the reflected light of the scanning laser radar. Recognition method. Repeated exploration in front of the vehicle with a scanning laser radar mounted on the vehicle,
Based on the exploration, the observation value of the vehicle index related to the reception state of the laser radar is detected,
Calculate the recognition evaluation value that changes depending on the difference between the observed value detected this time and the set reference value,
An object recognition method characterized by recognizing the object by calculating the probability of an object ahead of the host vehicle based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle.   When the observation value of the vehicle index related to the reception state of the scanning laser radar of reflected light by the object in front of the host vehicle is the reflected light reception intensity of the scanning laser radar, and the set reference value is the vehicle The object recognition method according to claim 6, wherein the reflected light received light intensity is an average value. Repeated exploration in front of the vehicle with a scanning laser radar mounted on the vehicle,
Based on the exploration, the laser radar detects the observation values of a plurality of vehicle indicators related to the position and size of an object in front of the vehicle where the reflected light is received, and the reception state of the reflected light of the laser radar,
Calculate each recognition evaluation value that changes depending on the difference between each observation value detected this time and the estimated value estimated, the previous detection value, and each comparison value that is set with each reference value,
An object recognition method, wherein the object is recognized by calculating the probability of the object based on a set proportional characteristic between an average value of the recognition evaluation values and a probability of being a vehicle. The observed value of the vehicle index related to the position of the object in front of the host vehicle is the distance from the host vehicle to the object in front of the host vehicle calculated from the transmission / reception time difference of the scanning laser radar, and the time of the object based on the time change of the distance Each relative speed,
The observed value of the vehicle index related to the size of the object is the width of the object detected from the interval of the light receiving positions of the reflected light of the scanning laser radar,
The observation value of the vehicle index related to the reception state of the reflected light is the reflected light reception intensity of the scanning laser radar,
The comparison value of the distance and the relative speed is composed of predicted values of the vehicle index for the distance and the relative speed, which are estimated by calculating from the change characteristics of the distance and the relative speed based on the already detected observation values. ,
The comparison value of the width comprises the observation value of the width detected last time,
9. The object recognition method according to claim 8, wherein the comparison value of the reflected light reception intensity comprises a reference value set to an average value of the reflected light reception intensity when the object is a vehicle. Normalize the recognition evaluation value of each observation value of the vehicle index so that the variation of each recognition evaluation value is substantially equal,
10. The probability of an object ahead of the host vehicle is calculated based on a set proportional characteristic between the normalized average value of the respective recognition evaluation values and the probability of being a vehicle. Object recognition method. A scanning laser radar that is mounted on the vehicle and repeatedly explores the front of the vehicle,
Based on the exploration, an index detection means for detecting an observation value of a vehicle index related to the position of an object ahead of the host vehicle where the laser radar receives reflected light;
A predicted value of the vehicle index is estimated from the observation value already detected by the index detection means, and the difference between the prediction value and the observation value detected by the index detection means at this time is higher or lower than the increase or decrease of the difference. A recognition evaluation value calculation means for calculating a changing recognition evaluation value;
Recognizing output means for calculating the probability of the object based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and outputting a calculation result as a recognition result of the object; Object recognition device.   A distance from the vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar is detected as an observation value of the vehicle index related to the position of the object ahead of the host vehicle by the index detection unit, and a recognition evaluation value calculation unit The object recognition apparatus according to claim 11, wherein a predicted value of the vehicle index is calculated and estimated from a change characteristic of the distance.   As an observation value of the vehicle index related to the position of the object ahead of the host vehicle by the index detection means, the relative of the object based on the time change of the distance from the host vehicle to the object calculated from the transmission / reception time difference of the scanning laser radar 12. The object recognition apparatus according to claim 11, wherein a speed is detected, and a predicted value of the vehicle index is calculated and estimated from a change characteristic of the relative speed by a recognition evaluation value calculation unit. A scanning laser radar that is mounted on the vehicle and repeatedly explores the front of the vehicle,
Based on the exploration, an index detection means for detecting an observation value of a vehicle index related to the size of an object ahead of the host vehicle where the laser radar receives reflected light; and
A recognition evaluation value calculating means for calculating a recognition evaluation value that changes in height in the opposite direction of increase or decrease of the difference from the difference between the observation value detected this time and the observation value detected last time by the index detection means;
Recognizing output means for calculating the probability of the object based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and outputting a calculation result as a recognition result of the object; Object recognition device.   The width of the object is detected from the interval of the reception position of the reflected light of the scanning laser radar as the observation value of the vehicle index related to the size of the object ahead of the host vehicle by the index detection means. The object recognition apparatus according to claim 14. A scanning laser radar that is mounted on the vehicle and repeatedly explores the front of the vehicle,
Based on the exploration, an index detection means for detecting an observation value of a vehicle index related to the reception state of the laser radar;
A recognition evaluation value calculating means for calculating a recognition evaluation value that changes depending on a difference between the observation value detected this time by the index detection means and a set reference value;
Recognizing output means for calculating the probability of an object ahead of the host vehicle based on a set proportional characteristic between the recognition evaluation value and the probability of being a vehicle, and outputting the calculation result as a recognition result of the object An object recognition device.   The index detection means detects the reflected light reception intensity of the scanning laser radar as an observation value of the vehicle index related to the reception state of the reflected light scanning laser radar, and the recognition evaluation value calculation means uses the reference value as the reference value The object recognition apparatus according to claim 16, wherein an average value of the reflected light reception intensity when the object is a vehicle is set. A scanning laser radar that is mounted on the host vehicle and repeatedly searches the front of the host vehicle,
An index for detecting observation values of a plurality of vehicle indices related to the position and size of an object in front of the vehicle where the laser radar receives reflected light and the reception state of the reflected light of the laser radar based on the exploration. Detection means;
Each recognition evaluation value that changes in level according to the difference between each observation value detected this time by the index detection means and each comparison value composed of the estimated value, the previous detection value, and the set reference value is calculated. A recognition evaluation value calculation means;
A recognition output means for calculating the probability of the object based on a set proportional characteristic between the average value of the recognition evaluation values and the probability of being a vehicle, and outputting the calculation result as a recognition result of the object; An object recognition device that recognizes the feature. As an observation value of the vehicle index related to the position of the object ahead of the host vehicle by the index detection means, the distance from the host vehicle to the object ahead of the host vehicle calculated from the transmission / reception time difference of the scanning laser radar, the time change of the distance Detecting the relative speed of the object based on the object, and detecting the lateral width of the object detected from the interval of the reception position of the reflected light of the laser radar as an observation value of the vehicle index related to the size of the object, As an observation value of the vehicle index related to the reception state of the reflected light of the laser radar, the reflected light reception intensity of the laser radar is detected,
As a comparison value of the distance and the relative speed, the recognition evaluation value calculation means calculates a predicted value of the vehicle index for each of the distance and the relative speed from the change characteristics of the distance and the relative speed based on the already detected observation values. And calculating the comparison value of the horizontal width based on the observation value of the horizontal width detected last time, and setting the comparison value of the reflected light reception intensity of the laser radar to the average value of the reflected light reception intensity when the object is a vehicle. 19. The object recognition apparatus according to claim 18, wherein the set reference value is used.   The recognition evaluation value calculation means normalizes the recognition evaluation value of each observation value of the vehicle index so that the variations of the recognition evaluation values are substantially equal, and the recognition output means averages the respective recognition evaluation values normalized. 20. The object recognition apparatus according to claim 18 or 19, wherein the probability of an object ahead of the host vehicle is calculated based on a set proportional characteristic between a value and a probability of being a vehicle.

Images