JP6645074B2 - Obstacle detection system - Google Patents

Description

  The present invention relates to an obstacle detection system.

  2. Description of the Related Art Conventionally, there is an object recognition device that detects an object by a sensor and determines whether the detected object is dangerous or not (for example, see Patent Document 1). The object recognition device of Patent Literature 1 is mounted on a vehicle, irradiates a laser beam, detects an object in front of the vehicle, and detects an object having a possibility of collision.

JP 2010-181246 A

  In the related art, for example, an object existing ahead of a preceding vehicle is excluded from an object to be determined as dangerous if it is defined as an erroneously detected object. Further, in the related art, a large number of target object candidates may be detected by the sensor. Therefore, it has not been easy to accurately recognize obstacles that are dangerous objects.

  An object of the present invention is to provide an obstacle detection system capable of accurately detecting an obstacle.

  An obstacle detection system according to the present invention includes: a detection unit that detects an object; and a confidence setting unit that sets a certainty factor that indicates a possibility that the object is an obstacle based on detection data of the object acquired by the detection unit. A degree setting unit, a tracking processing unit that performs a tracking process of determining whether a plurality of pieces of detection data having different detection times are for the same target object, and the plurality of pieces of detection data are the same in the tracking processing unit. A certainty update unit that updates the certainty factor of the target object using the detection data related to the same target object when it is determined that the target object is related to the target object.

  This obstacle detection system performs tracking processing and updates the certainty factor of the target object when it is determined that the plurality of detection data items belong to the same target object. Thus, the degree of certainty can be appropriately updated. Thereby, an obstacle can be detected with high accuracy according to the updated certainty factor.

  In addition, the detection data of the object includes information regarding the brightness of the object, and the confidence setting unit compares the brightness difference between the object and the surroundings of the object with the case where the brightness difference is small. Therefore, the confidence can be set high. When the brightness difference between the object and its surroundings is large, the possibility of erroneous detection is lower than when the brightness difference between the object and its surroundings is small, so the brightness difference between the object and its surroundings is small. Is large, the certainty of the object is set high, and when the brightness difference between the object and its surroundings is small, the certainty of the object is set low, so that obstacles are detected with high accuracy.

  Further, the detection data of the object includes information on the hue of the object, and the certainty factor setting unit compares the hue difference between the object and the periphery of the object with the hue difference when the hue difference is small. Therefore, the confidence can be set high. When the hue difference between the object and its surroundings is large, the possibility of erroneous detection is lower than when the hue difference between the object and its surroundings is small, so the hue difference between the object and its surroundings is small. Is large, the hue difference between the object and its surroundings is set high, and when the hue difference between the object and its surroundings is small, the certainty of the object is set low, so that obstacles are accurately detected.

  In addition, the detection data of the object includes information on the external dimensions of the object, and the confidence setting unit compares the external dimensions of the object with the confidence level when the external dimensions of the object are large compared to when the external dimensions of the object are small. Can be set higher. When the object is large, the possibility of erroneous detection is lower than when the object is small.Therefore, when the object is large, the confidence of the object is set high, and when the object is small, By setting the certainty factor of the object low, obstacles can be accurately detected.

  Further, the detection data of the object includes information on the distance from the detection unit to the object, and the certainty factor setting unit, when the distance to the object is short, as compared with the case where the distance to the object is long, Confidence can be set high. An object detected at a position closer to the detection unit is less likely to be erroneously detected than an object detected at a position farther from the detection unit. Obstacles can be detected with high accuracy by setting the certainty factor of the object high when the distance is short and by setting the low certainty factor when the distance to the object is long.

  In addition, the configuration further includes an obstacle determining unit that determines whether the updated certainty is equal to or greater than the determination threshold and determines that the target object is an obstacle when the certainty is equal to or greater than the determination threshold. Good. This prevents the target from being detected as an obstacle when the certainty is less than the determination threshold.

  Further, the certainty update unit may be configured to calculate a weighted average of a plurality of pieces of detection data and update the certainty of the target object. The certainty factor of the target object can be appropriately updated using the weighted average for the plurality of pieces of detection data.

  Further, the certainty update unit may include a plurality of types of detection units, and the certainty update unit may change the weight in the weighted average according to the type of the detection unit to update the certainty of the object. Good. This makes it possible to appropriately update the certainty factor of the target object in consideration of the detection accuracy that varies depending on the type of the detection unit.

  Also, the certainty factor updating unit may be configured to reduce the certainty factor of the target object when the tracking processing unit does not determine the tracking of the target object. Accordingly, if the target object detected by the detection unit is not detected thereafter, the tracking determination is not performed, so that the degree of certainty gradually decreases. Therefore, the possibility that the undetected target is detected as an obstacle is reduced.

  According to the obstacle detection system of the present invention, an obstacle can be accurately detected.

1 is a perspective view illustrating a state in which a vehicle equipped with an obstacle detection device according to an embodiment of the present invention travels on a highway. 1 is a block diagram illustrating an obstacle detection device according to an embodiment of the present invention. It is a graph which shows the example of a change of a certainty factor. It is a graph which shows the example of a change of a certainty factor. It is a flowchart which shows the processing procedure in an image processing part and a 3DLR processing part. 9 is a flowchart illustrating a processing procedure in an integration processing unit.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In each of the drawings, the same or corresponding portions have the same reference characters allotted, and overlapping description will be omitted.

  The obstacle detection device 1 shown in FIGS. 1 and 2 is mounted on, for example, a vehicle M and detects an object (for example, a falling object B) in front of the vehicle. The obstacle detection device 1 includes a camera 2, 3DLR (three-dimensional laser radar) 3, a wheel speed sensor 4, and an obstacle detection unit 5.

  The obstacle detection device 1 tracks the detected object, and determines an object having a high degree of certainty among the tracked objects as an obstacle. The “obstacle” is an object that may obstruct the traveling of the vehicle, and it is necessary to change the traveling so as to avoid the object, or to stop and remove the object. The “certainty factor” is an index indicating the likelihood that the target object is an obstacle. When the degree of certainty is high, the possibility that the object is an obstacle is high, and when the degree of certainty is low, the possibility that the object is an obstacle is low.

  The camera 2 (detection unit) captures an image of the front of the vehicle and acquires image information (detection data) of a road, an object, and the like. Image information obtained by the camera 2 is input to the obstacle detection unit 5.

  The 3DLR (detection unit) 3 irradiates a laser in front of the vehicle and receives a reflected laser reflected by an object on the road. Information (detection data) regarding the reflected laser received by the 3DLR 3 is input to the obstacle detection unit 5.

  The wheel speed sensor 4 is a sensor that acquires information on the rotation angle of the wheel of the vehicle M. The wheel speed sensor 4 includes, for example, an encoder, and the encoder measures a rotation pulse of the wheel. A signal related to the wheel rotation pulse is input to the obstacle detection unit 5.

  The obstacle detection unit 5 includes an image processing unit 6, a 3DLR processing unit 7, a linear calculation unit 8, an integrated processing unit 9, and a system control unit 10. The obstacle detection unit 5 includes a CPU that performs arithmetic processing, ROM and RAM serving as storage units, an input signal circuit, an output signal circuit, a power supply circuit, and the like.

  The system control unit 10 controls the overall operation of the obstacle detection unit 5, and manages the generation of blocks and the suspension of functions of the image processing unit 6, 3DLR processing unit 7, linear calculation unit 8, and integration processing unit 9. Further, the system control unit 10 manages input and output of data with a display unit 15, an operation unit 16, and a vehicle control unit 17 described below.

  The image processing unit 6 performs image processing on the sensor data input from the camera 2. The image processing unit 6 performs, for example, binarization processing as image processing, and detects information on the target object and information on the white line of the road. The sensor data input from the camera 2 includes information on the hue of the object, information on the brightness of the object, information on the size (external dimensions) of the object, and information on the position of the object.

  In addition, the image processing unit 6 functions as a certainty setting unit that sets certainty for the detected object. Based on the sensor data, the image processing unit 6 determines the position of the object (the position in the horizontal direction, the position in the traveling direction of the vehicle), the size of the object (the size in the width direction, the size in the height direction), Calculate hue and lightness. Further, the image processing unit 6 calculates the hue and lightness around the object.

  The image processing unit 6 determines the hue of the object and the hue around the object based on the sensor data of the object, and when the hue difference between the object and the periphery of the object is large, the hue difference Is set higher than the case where is small. The image processing unit 6 sets the certainty higher when the average hue of the detected target object is determined to be higher than when the average hue is determined to be lower.

  The image processing unit 6 determines the brightness of the target object and the brightness around the target object based on the sensor data of the target object, and when the brightness difference between the target object and the surrounding of the target object is large, the brightness difference Is set higher than the case where is small. The image processing unit 6 sets the certainty higher when the average brightness of the detected target object is determined to be higher than when the average brightness is determined to be lower.

  The image processing unit 6 determines the size of the target object based on the sensor data of the target object. When the size of the target object (actual size) is large, the certainty factor is compared with the case where the size of the target object is small. Set high. The image processing unit 6 calculates and determines the actual size of the target object by comparing camera parameters such as the angle of view and linear information of the path with the pixel size in the image.

  The image processing unit 6 determines the position of the target based on the sensor data of the target, and compares the position of the target with the vehicle M when the distance between the vehicle M and the target is long. To increase the confidence.

The image processing unit 6 calculates the certainty factor using, for example, the following equation (1).
C ^CAM = f (Intensity, Size, Dist) (1)
The ^CAM is a certainty factor based on the sensor data acquired by the camera 2. In Equation (1), “Intensity” is the hue or lightness of the object, “Size” is the external dimension of the object, “Dist” is the distance between the vehicle and the object, and f is the confidence. This is a function for calculation.

  The 3DLR processing unit 7 performs data processing on the sensor data input from the 3DLR 3. The sensor data input from the 3DLR includes the position of the object (the position in the horizontal direction, the position in the traveling direction of the vehicle, the position in the vertical direction), the size of the object (the size in the width direction, the size in the height direction, (The size in the depth direction), the reflection intensity of the laser reflected on the object, the number of points of the laser reflected on the object (measurement points), the position of the center of gravity of the point group that is a plurality of points on which the laser was irradiated (the center of gravity in the horizontal direction, vehicle) , The center of gravity in the traveling direction, the center of gravity in the vertical direction), and information on the dispersion of the point cloud.

  The 3DLR processing unit 7 functions as a certainty setting unit that sets certainty for the object detected by the 3DLR3. Based on the sensor data, the 3DLR processing unit 7 determines the position of the object (the position in the horizontal direction, the position in the traveling direction of the vehicle, the position in the vertical direction), the size of the object (the size in the width direction, the size in the height direction). , The size in the depth direction), the reflection intensity of the laser reflected on the object, the number of points of the laser reflected on the object (the number of measurement points), the center of gravity of the point group that is a plurality of points on which the laser was irradiated (the center of gravity in the horizontal direction, The center of gravity in the traveling direction of the vehicle, the center of gravity in the vertical direction, and the variance of the point cloud are calculated.

  The 3DLR processing unit 7 determines the size of the target object based on the sensor data of the target object, and when the size of the target object (actual size) is large, the degree of certainty is compared with the case where the size of the target object is small. Set high.

  The 3DLR processing unit 7 determines, based on the sensor data of the target object, the reflection intensity of the laser beam reflected on the target object, and when the laser reflection intensity is high, the confidence level is compared with the case where the laser reflection intensity is low. Set high.

  The 3DLR processing unit 7 determines the number of lasers reflected on the object based on the sensor data of the object, and when the number of lasers reflected on the object is large, the number of lasers reflected on the object is small. The confidence is set higher than in the case.

  The 3DLR processing unit 7 determines the position of the center of gravity of the point group hit by the plurality of lasers based on the sensor data of the object, and determines the position of the center of gravity of the point group when the position of the center of gravity of the point group is close to the vehicle. The certainty factor is set higher than when the vehicle is far from the vehicle. Further, the 3DLR processing unit 7 sets a certainty factor higher when the position of the center of gravity of the point cloud is closer to the center of the route than when the position of the center of gravity of the point cloud is farther from the center of the route.

  The 3DLR processing unit 7 determines the distribution of the point group based on the sensor data of the object, and when the size of the confidence interval of the variance of the point group is equal to the outer size of the object, the 3DLR processing unit 7 determines the variance of the point group. Is set higher than the case where the size of the confidence interval is significantly different from the external size of the object.

The 3DLR processing unit 7 calculates the certainty factor using, for example, the following equation (2).
C ^3DLR = g (Num of Spots, Trace) (2)
In addition, ^C3DLR is a certainty factor based on the sensor data acquired by 3DLR3. In Equation (2), “Num of Spots” is the number of reflected lasers, “Trace” is a coefficient indicating the tracking success / failure result determined by the 3DLR 3 itself, and g is a function for calculating the certainty factor. It is. The processing result by the 3DLR processing unit 7 is output to the integration processing unit 9.

  The linear calculation unit 8 calculates the travel distance of the vehicle using the linear information of the route stored in the storage unit and the vehicle speed acquired by the wheel speed sensor 4. "Route linear information" is information on the three-dimensional shape of the route corresponding to the distance from the position serving as the starting point of the route, information on the horizontal curvature of the curve on the road, and information on the vertical gradient on the slope. Is included. The linear calculation unit 8 calculates own vehicle position information indicating a position from the starting point of the vehicle based on the linear information of the route and the vehicle speed.

  The integration processing unit 9 performs an integration process of integrating sensor data of an object acquired from a plurality of sensors. The integration processing unit 9 integrates a plurality of temporally different sensor data as temporally continuous data in order to obtain data suitable for the tracking processing and the obstacle determination processing.

  The integration processing unit 9 includes a tracking processing unit 11, a certainty factor updating unit 12, an obstacle determining unit 13, and a mask processing unit 14.

  The tracking processing unit 11 performs tracking processing (tracking) of the target object using, for example, a Kalman filter. A tracking process for associating sensor data relating to the same object between data frames that are temporally preceding and succeeding is performed. Thereby, it is possible to recognize the object once detected with time. The tracking processing unit 11 performs a tracking process for each type of sensor. The tracking processing unit 11 separately performs a tracking process using the sensor data acquired by the camera 2 and a tracking process using the sensor data acquired by the 3DLR 3.

The credibility update unit 12 updates the credibility based on the sensor data when the tracking processing unit 11 performs the tracking determination. For example, the sensor data _{D S1} at time t = 1, the case where the sensor data _{D S2} at time t = 2 is tracked determined to be the same object (if matched), the sensor data _{D S1, D S2} Is updated on the basis of the certainty factor associated with. For example, the confidence may be updated by averaging the confidences associated with the sensor data D _S1 and D _S2 , or the confidence may be updated using a weighted average.

The certainty updating unit 12 calculates a weighted average using the following equation (3) and updates the certainty of the target object.
_{C k + 1 = wC k +} w s C k S ... (3)
Here, C _{k + 1} is the certainty at time k + 1, C _k is the certainty at time k, C _k ^S is the certainty set based on the sensor data of (1) or (2), and w is a weighting factor for the current value, w _s is the weighting factor for measured values (weighting factor according to the type of sensor).

Further, the weight coefficient w _s for confidence C _S based on the sensor data from the camera 2 may be the same value and the weight coefficient w _s for confidence C _S based on the sensor data by 3DLR3, the value may be different. For example, when the detection accuracy of the sensor is high, the weight coefficient according to the sensor type is set higher than when the detection accuracy of the sensor is low.

  Further, when the tracking processing unit 11 cannot determine the tracking of the target, the certainty updating unit 12 updates the certainty so as to reduce the certainty of the target. For example, the current certainty may be reduced at a fixed rate, or a certain value may be subtracted.

  3 and 4 are graphs showing examples of changes in the certainty factor. 3 and 4, the horizontal axis indicates the parameters of the sensor data (values related to hue, lightness, size, or distance), and the vertical axes indicate the certainty factor. FIG. 3A shows a case where the certainty factor is constant even when the parameter increases. FIG. 3B shows a case where the certainty factor increases in proportion to an increase in the parameter (linear change).

  FIG. 3C shows a case where the certainty increases with an increase in the parameter within the range of the upper limit and the lower limit of the certainty. For example, if the parameter is equal to or less than P1, the certainty factor is constant at the lower limit. If the parameter is greater than or equal to P1 and less than or equal to P2, the confidence increases in proportion to the parameter increase (linear change). If the parameter is equal to or greater than P2, the certainty factor is constant at the upper limit.

  FIG. 3D shows a case where the certainty factor nonlinearly changes and increases as the parameter increases. For example, the rate of increase in the certainty factor increases with an increase in the parameter.

  FIG. 4A shows a case where the certainty increases with an increase in the parameters within the range of the upper limit and the lower limit of the certainty. For example, if the parameter is equal to or less than P1, the certainty factor is constant at the lower limit. When the parameter is equal to or more than P1 and equal to or less than P2, the certainty factor increases as the parameter increases (nonlinear change). For example, the rate of increase of the certainty factor in the non-linear change section increases with an increase in the parameter. If the parameter is equal to or greater than P2, the certainty factor is constant at the upper limit.

  FIG. 4B shows a case where the certainty factor increases nonlinearly as the parameter increases. For example, the rate of increase in certainty decreases with an increase in parameters.

  FIG. 4C shows a case where the certainty factor changes stepwise and increases with an increase in the parameter. For example, when the parameter is not less than P11 and less than P12, the certainty is constant at C11, and when the parameter is not less than P12 and less than P13, the certainty is constant at C12 (C11 <C12).

  In the case shown in FIG. 4D, the degree of certainty differs depending on the distance from the center in the width direction of the traveling path to the object. For example, a plurality (for example, four) models of the change in the certainty factor are set so that the certainty factor increases in proportion to the increase in the parameter. When the distance from the center in the width direction of the traveling path to the object is short, the confidence is higher and the confidence is higher than when the distance from the center in the width direction of the traveling path to the object is longer. The rate of increase in degrees is high.

  When an object detected based on the sensor data satisfies a predetermined condition, the mask processing unit 14 determines that the object is not an obstacle. Thereby, the calculation load on the integration processing unit 9 is reduced. As the predetermined condition, for example, a condition that can determine that the detected object has a low degree of risk and does not affect the running of the vehicle is set.

  The obstacle determining unit 13 determines whether the object detected by the camera 2 or the 3DLR 3 is an obstacle based on the certainty factor. The obstacle determining unit 13 determines that the object detected by the sensor is an obstacle when the certainty factor is equal to or greater than the threshold. The processing result by the integration processing unit 9 is output to the system control unit 10.

  The display unit 15 and the operation unit 16 are electrically connected to the obstacle detection unit 5. The display unit 15 is, for example, a liquid crystal display device, and displays an image acquired by the camera 2. Further, based on the information output from the obstacle detection unit 5, the driver is notified by highlighting the object which is the obstacle.

  The operation unit 16 is, for example, a touch panel of a liquid crystal display device, and outputs a signal based on a driver's operation input to the system control unit 10. For example, the driver operates the operation unit 16 when performing an input operation on an object that has been confirmed as not an obstacle.

  Further, a vehicle control unit 17 that controls the vehicle M is electrically connected to the obstacle detection device 1. For example, an alarm 18 and a brake 19 are electrically connected to the vehicle control unit 17. The vehicle control unit 17 can control the alarm device 18 based on a signal from the system control unit 10 to emit an alarm sound. Further, the vehicle control unit 17 can operate the brake 19 based on a signal from the system control unit 10 to decelerate the vehicle.

  Next, the operation of the obstacle detection device 1 will be described. This will be described with reference to the flowcharts shown in FIGS. FIG. 5 is a flowchart illustrating a processing procedure in the image processing unit 6 and the 3DLR processing unit 7. FIG. 6 is a flowchart illustrating a processing procedure in the integration processing unit 9.

  First, the setting of the certainty factor in the obstacle detection unit 5 will be described with reference to the flowchart in FIG. The process of setting the certainty factor shown in FIG. 5 (steps S1 to S4) is executed separately in the image processing unit 6 and the 3DLR processing unit 7.

  Here, after the processing in the image processing unit 6 is described, the processing in the 3DLR processing unit 7 will be described.

  The image processing unit 6 acquires image information as sensor data acquired by the camera 2 (Step S1). The sensor data includes information on the hue, brightness, size, and position of the object. Note that the required detection performance of the plurality of sensors (cameras 2 and 3DLR3) is as follows. For example, "at night, when the object is stationary, it exists in a section 50 m or more and 450 m or less in front of the vehicle, has a reflectance of about a standard value, and can detect an object having a soccer ball or a small size of about a small Boston back. ".

  Next, the image processing unit 6 acquires the vehicle position information from the linear calculation unit 8 (Step S2). The linear calculation unit 8 calculates the vehicle speed of the vehicle M from the wheel speed sensor 4, calculates the moving distance of the vehicle M from the starting point based on the vehicle speed and the linear information of the route, and calculates the own vehicle position information. . The calculated vehicle position information is output to the image processing unit 6.

  Subsequently, the image processing unit 6 sets the certainty factor of the target based on the sensor data and the vehicle position information (Step S3). As described above, the image processing unit 6 determines the hue of the target, the brightness of the target, the size of the target, and the distance from the vehicle M to the target, and sets the certainty of the target. For example, the image processing unit 6 may calculate the certainty using a mathematical expression for calculating the certainty, or may calculate the certainty using a map. In the map for calculating the certainty factor, the hue of the target object, the brightness of the target object, the size of the target object, and the distance to the target object are associated with the certainty factor. The map for calculating the certainty factor is stored in, for example, the storage unit of the obstacle detection unit 5.

  Then, the image processing unit 6 outputs sensor data and data on the set certainty factor to the integration processing unit 9 as various data (step S4).

  Next, processing in the 3DLR processing unit 7 will be described. In step S1, the 3DLR processing unit 7 acquires information on the laser beam reflected on the object as the sensor data acquired by the 3DLR3. In this sensor data, the position of the object, the size of the object, the reflection intensity of the laser reflected on the object, the number of lasers reflected on the object, the center of gravity of a plurality of points that the laser hit, Contains information about the variance of the point cloud.

  Next, the 3DLR processing unit 7 proceeds to step S2, acquires the vehicle position information from the linear calculation unit 8, and then proceeds to step S3 to determine the degree of certainty of the target object based on the sensor data and the vehicle position information. Set.

  As described above, the 3DLR processing unit 7 determines the size of the target object, the distance from the vehicle M to the target object, the reflection intensity of the laser reflected on the target object, the number of lasers reflected on the target object, and the number of lasers reflected on the target object. The position of the center of gravity of the point group, which is a point, and the variance of the point group are determined, and the certainty factor of the object is set. The 3DLR processing unit 7 may calculate the certainty using, for example, a mathematical expression for calculating the certainty, or may calculate the certainty using a map.

  Then, the image processing unit 6 outputs sensor data and data on the set certainty factor to the integration processing unit 9 as various data (step S4).

  Next, a processing procedure in the integration processing unit 9 will be described with reference to a flowchart of FIG.

  First, the integration processing unit 9 inputs various data output from the image processing unit 6 and the 3DLR processing unit 7 (Step S11). The integration processing unit 9 integrates various data as appropriate.

  Subsequently, in step S12, the tracking processing unit 11 of the integration processing unit 9 determines the sensor (camera 2 or 3DLR3) that has acquired the sensor data. The sensor data includes an identification signal indicating the acquired sensor. The tracking processing unit 11 reads this identification signal and determines the sensor that has acquired the sensor data.

  Next, in step S13, the tracking processing unit 11 performs a tracking process (tracking calculation) for each sensor using the sensor data acquired by the sensor. The tracking processing unit 11 associates sensor data relating to the same object between data frames that are temporally successive. If the presence of the same object can be confirmed between a plurality of data frames, it is determined that the object has been tracked (tracking determination).

  The tracking processing unit 11 performs tracking calculation of, for example, about 100 objects at a cycle of 100 ms. Further, the tracking processing unit 11 assumes that the target object is stationary with respect to the ground, and that a change in observation by the sensor depends only on the motion of the vehicle. Further, in order to reduce the calculation load, the tracking processing unit 11 does not calculate the correlation between the positions of the objects, but performs the tracking calculation for each object. For example, in the calculation method of the conventional tracking processing, when an error due to an erroneous match occurs, the error is distributed to the tracking processing for a plurality of objects via the Kalman gain. However, the tracking processing unit 11 does not calculate the correlation between the positions of the elephants, but performs tracking processing individually for the objects, so that the influence of errors on the tracking processing of other objects is prevented.

  If the object can be tracked (if the tracking can be determined, S14; YES), the process proceeds to step S15. If the object cannot be tracked (if the tracking cannot be determined, S14; NO), the process proceeds to step S15. Proceed to S16.

  In step S15, the certainty factor updating unit 12 updates the certainty factor using the certainty factor associated with the plurality of sensor data. Here, the certainty factor of the target object is updated using the certainty factor linked to the plurality of sensor data associated with the tracking determination of the target object. The certainty updating unit 12 updates the certainty associated with the plurality of sensor data by setting a new certainty using, for example, a weighted average.

  In step S16, the certainty factor updating unit 12 updates the certainty factor so as to reduce the certainty factor of the target object for which the tracking determination has not been performed. For example, the current certainty may be reduced at a fixed rate, or a certain value may be subtracted.

  Next, in step S17, the obstacle determining unit 13 determines whether or not the certainty factor of the target is equal to or greater than a determination threshold. If the certainty is equal to or more than the determination threshold (step S17; YES), the object is determined to be an obstacle, and the process proceeds to step S18. If the certainty is less than the determination threshold (step S17; NO), Ends the processing here.

  In addition, when the target object satisfies the predetermined condition, the mask processing unit 14 can determine that the target object is not an obstacle. Accordingly, the determination process by the obstacle determination unit 13 can be avoided, and unnecessary warnings and avoidance operations can be suppressed, so that the load on the obstacle detection unit 5 and the vehicle control unit 17 can be reduced. Can be.

  Subsequently, in step S18, the integration processing unit 9 outputs the determination result by the obstacle determination unit 13 to the system control unit 10, and the system control unit 10 determines that the system control unit 10 is an obstacle in step S17. Is notified to the driver. Specifically, when displaying an image by the camera 2 on the display unit 15, the corresponding object is highlighted (warned) as an obstacle. As the highlighting, for example, an obstacle is surrounded by a red frame. Further, a numerical value of the certainty factor may be displayed near the target object.

  When the driver is notified that the vehicle is an obstacle, the driver checks the obstacle visually and stops the vehicle or operates the steering wheel when the driver confirms that the vehicle is actually an obstacle. By changing the course, contact with obstacles can be avoided.

  If the driver visually determines that the object is not an obstacle, the driver performs an operation input using the operation unit 16 and sets the object as a canceled object (an object that is not an obstacle).

  After executing the processing in step S18, the obstacle detection unit 5 ends the processing in FIG. 6, returns to step S1, and repeats the same processing.

  According to the obstacle detection device 1, when the tracking process is performed and the tracking determination can be performed, the certainty factor of the target object is updated. Therefore, the certainty factor is appropriately updated using a plurality of sensor data having different detection times. be able to. Thereby, an obstacle can be detected with high accuracy according to the updated certainty factor.

  In addition, when the tracking processing unit 11 does not determine the tracking of the target, the certainty updating unit 12 updates the certainty of the target so as to reduce the certainty of the target. As a result, if the object detected by the camera 2 or the 3DLR 3 is not subsequently detected, the tracking determination is not performed, so that the certainty degree gradually decreases. Therefore, the possibility that the undetected target is detected as an obstacle is reduced. As a result, it is possible to reduce the possibility that an object having a high possibility of erroneous detection is detected as an obstacle.

  Further, the obstacle detection device 1 can set a certainty factor according to the brightness difference between the target object and its surroundings, and can update the certainty factor by using the set certainty factor. Accordingly, by updating the certainty factor of the target object in consideration of the data regarding the brightness of the target object included in the sensor data, erroneous detection can be reduced and the accuracy of detecting the obstacle can be improved.

  Further, the obstacle detection device 1 can set the certainty according to the hue difference between the target object and the surroundings, and update the certainty using the set certainty. Accordingly, by updating the certainty factor of the target object in consideration of the data regarding the hue of the target object included in the sensor data, erroneous detection can be reduced and the accuracy of detecting the obstacle can be improved.

  In the obstacle detection device 1, the certainty factor can be set according to the external size of the object, and the certainty factor can be updated using the set certainty factor. Accordingly, by updating the certainty factor of the target object in consideration of the data regarding the outer shape size of the target object included in the sensor data, it is possible to reduce erroneous detection and improve the accuracy of detecting an obstacle. .

  Further, the obstacle detection device 1 can set a certainty factor according to the distance from the vehicle M to the target, and update the certainty factor using the set certainty factor. Accordingly, by updating the certainty factor of the target object in consideration of the data on the distance to the target object included in the sensor data, it is possible to reduce erroneous detection and improve the accuracy of detecting an obstacle. .

  Further, when the distance from the vehicle M to the obstacle is short, the time for avoiding the obstacle and the time for confirming the obstacle are short, so it is necessary to take immediate measures. In the obstacle detection device 1, when the distance from the vehicle M to the object is short, the certainty of the object is set to be high to increase the possibility that the object is detected as an obstacle, and Can be notified quickly and reliably. As a result, the driver can speed up the braking operation and reliably avoid the collision with the obstacle.

  In addition, the certainty factor updating unit 12 updates the certainty factor of the target object by calculating a weighted average of a plurality of sensor data according to the type of the sensor, and reflects the importance of the sensor data based on the type of the sensor. Thus, the certainty factor of the object can be updated appropriately. Thereby, erroneous detection can be further reduced, and the accuracy of detecting obstacles can be improved.

  In addition, the obstacle determining unit 13 determines that the object is an obstacle when the certainty is equal to or greater than the determination threshold. Therefore, when the certainty is less than the determination threshold, the obstacle is detected as an obstacle. Is prevented. As a result, the accuracy of detecting obstacles can be improved.

  Further, in the object recognition device of the related art, a large number of target object candidates tend to be detected by a sensor in order to satisfy a required specification. Further, in the conventional object recognition device, a large number of detection results may be obtained in order to determine whether or not the object is a dangerous target using a single detection result of the sensor. In the obstacle detection device 1 of the present embodiment, the confidence level can be updated by integrating a plurality of pieces of detection data, so that the number of data used in the calculation in the obstacle determination unit 13 can be reduced. The obstacle detection device 1 can increase the accuracy (accuracy) of detecting an obstacle by using a plurality of pieces of detection data for one target object.

  The present invention is not limited to the embodiments described above, and various modifications as described below are possible without departing from the gist of the present invention.

  In the above-described embodiment, the camera is configured to include the camera 2 and the 3DLR 3 as the sensor. However, the sensor is not limited thereto. For example, a configuration including a plurality of types of cameras (a long-distance camera, a short-distance camera) having different detection ranges may be used. . Further, in the above embodiment, a configuration including a plurality of types of sensors is employed, but a configuration including a single type of sensor may be employed.

  Further, in the above embodiment, the certainty is set based on the information on the brightness of the object, the information on the hue of the object, the information on the external dimensions of the object, and the information on the distance to the object. The certainty may be set based on only the information, only the information related to the hue, only the information related to the outer dimensions, or only the information related to the distance to the object, or the certainty may be set using a plurality of pieces of information.

  In addition, the detection data of the object includes information regarding the brightness of the object, and the certainty factor setting unit determines the certainty factor when the brightness of the object is higher than when the brightness of the object is lower. May be set higher. When the object is bright, the possibility of erroneous detection is lower than when the object is dark.Therefore, when the object is bright, the confidence level of the object is set high, and when the object is dark, By setting the certainty factor of the object low, obstacles can be detected with high accuracy.

  In the above embodiment, the certainty factor updating unit 12 updates the certainty factor of the target object using the weighted average. However, the certainty factor may be updated without using the weighted average. For example, based on a plurality of certainty factors, the certainty factors may be updated using other formulas or maps.

  In the above-described embodiment, the certainty factor update unit 12 updates the certainty factor of the target object by changing the weight in the weighted average according to the type of the sensor. The weight in the weighted average may be changed according to conditions. For example, the weight in the weighted average may be changed based on the probability that the object exists at the place according to the type of the target object. When the existence probability of the object is high, the weight coefficient is increased as compared with the case where the existence probability of the object is low. This makes it possible to change the certainty factor of the target object in consideration of the existence probability of the target object. Note that the obstacle detection device 1 may determine the type of the target object based on the sensor data. Further, data on the existence probability of the object is stored in, for example, the storage unit of the integration processing unit 9.

  Further, in the above embodiment, when the tracking processing unit 11 cannot perform the tracking determination (Step S14; NO), the confidence is reduced (Step S16), but the confidence is maintained without decreasing the confidence. May be.

  In the above embodiment, the case where the obstacle detection system is mounted on the vehicle is described. However, the obstacle detection system may be mounted on a moving body other than the vehicle. For example, the obstacle detection system may be mounted on a robot, an airplane, a ship, a railway, or the like. Further, the obstacle detection system may be installed in another place or object without being mounted on the moving body. For example, an obstacle detection system may be installed in a factory, a store, a hospital, an airport, a station, a building around a traveling road, or the like, and a potentially dangerous object may be detected as an obstacle. The target object is not limited to a stationary object, but may be a moving object, a person, or an animal. In addition, the obstacle is not limited to a dangerous object, and may be an obstacle or an object that meets specific conditions.

1 Obstacle detection device (obstacle detection system)
2 Camera (detection unit)
3 3DLR (detection unit)
5 Obstacle detection unit 6 Image processing unit (confidence level setting unit)
7 3DLR processing unit (confidence level setting unit)
9 Integrated processing unit 11 Tracking processing unit 12 Confidence updating unit 13 Obstacle determination unit B Falling object (object, obstacle)

Claims (8)

A detection unit for detecting an object;
Based on the detection data of the target object obtained by the detection unit, a certainty factor setting unit that sets a certainty factor indicating the possibility that the target object is an obstacle,
A tracking processing unit that performs a tracking process to determine whether the plurality of pieces of detection data having different detection times are for the same object,
In the tracking processing unit, when it is determined that the plurality of the detection data is for the same object, using the detection data related to the same object, the certainty factor of the object And a confidence update unit for updating .
The detection unit includes a plurality of types of detection units including a three-dimensional laser radar,
The detection data of the object includes the number of lasers reflected on the object, the position of the center of gravity of a point group that is a plurality of points hit by the laser, and information on at least one of the variance of the point group,
The certainty factor updating unit sets each certainty factor based on the detection data of the target object obtained by each detecting unit, and calculates a weighted average using the certainty factors to obtain the target object. the obstacle detection system to update the confidence of. The detection data of the object includes information on the brightness of the object,
The obstacle according to claim 1, wherein the certainty factor setting unit sets the certainty factor higher when the brightness difference between the target object and the surroundings of the target object is larger than when the brightness difference is smaller. Object detection system. The detection data of the object includes information on a hue of the object,
The said certainty factor setting part sets the said certainty factor high when the hue difference between the said target object and the circumference | surroundings of the said target object is large compared with the case where the said hue difference is small. Obstacle detection system. The detection data of the object includes information on the external dimensions of the object,
The confidence factor setting unit according to any one of claims 1 to 3, wherein when the external dimensions of the object are large, the confidence is set higher than when the external dimensions of the object are small. Obstacle detection system as described. The detection data of the object includes information on a distance from the detection unit to the object,
5. The confidence factor setting unit according to claim 1, wherein when the distance to the target object is short, the confidence factor is set higher than when the distance to the target object is long. Obstacle detection system as described.   Determining whether the updated certainty factor is equal to or greater than a determination threshold, and further comprising, when the certainty factor is equal to or greater than the determination threshold, an obstacle determining unit that determines that the object is the obstacle. The obstacle detection system according to any one of claims 1 to 5. The said certainty factor update part changes the weight in the said weighted average according to the classification of the said detection part, and updates the said certainty factor of the said target object in any one of Claims 1-6 . Obstacle detection system. The obstacle detection according to any one of claims 1 to 7 , wherein the certainty update unit reduces the certainty of the object when the tracking processing unit does not determine the tracking of the object. system.

Images