Object recognition device, object recognition method, and object recognition program

Abstract

To further improve an accuracy of object recognition around a vehicle.SOLUTION: An object recognition device (6) comprises a sensor information acquisition unit (601), an object detection unit (602), an integration processing unit (604), and a recognition processing unit (605). The integration processing unit generates a detection performance curve for each of multiple types of object detection sensors, and then uses the detection performance curve to calculate a level of reliability of an object detection result, and integrates the calculated levels of reliability of the object detection results obtained for each of the multiple types of object detection sensors. The detection performance curve is generated based on the surrounding environment of the vehicle and / or feedback information regarding the object recognition results from the recognition processing unit. The recognition processing unit performs object recognition based on the integrated result of the levels of reliability determined by the integration processing unit.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to an object recognition device, an object recognition method, and an object recognition program to be mounted on a vehicle. [Background technology]

[0002] Various devices that perform so-called sensor fusion processing to recognize objects around a vehicle have been known. For example, Patent Document 1 discloses an external environment recognition device that recognizes the external environment of a vehicle equipped with multiple types of sensors whose detection ranges at least partially overlap, and includes an acquisition unit, a setting unit, and a generation unit. The acquisition unit acquires detection information from a camera, a laser radar, and a radio wave radar. The laser radar may also be referred to as LiDAR. LiDAR stands for Light Detection and Ranging or Laser Imaging Detection and Ranging. The setting unit assigns weights to each piece of detection information related to the overlapping parts of the detection ranges according to the driving scene of the vehicle. The generation unit integrates each piece of detection information with the assigned weights to generate recognition information of the external environment. This configuration enables more accurate recognition of the external environment through sensor fusion processing.

[0003] Specifically, in a sidewall curve scene in which the vehicle travels through a curved section on a road with sidewalls, the setting unit sets the weight of information detected by the radio wave radar to the smallest value and the weight of information detected by the laser radar to a value greater than that of information detected by the camera. Furthermore, during a weight adjustment period after the vehicle has entered a dark area from a bright area, the setting unit sets the weight of information detected by the laser radar to a value greater than that of information detected by the camera. Meanwhile, in a bright area entry scene immediately after the vehicle has entered a dark area from a bright area, the setting unit sets the weight of information detected by the camera to a value less than that of information detected by the laser radar. Furthermore, in a snowfall scene, the setting unit sets the weight of information detected by the camera to a value less than that of information detected by the laser radar and sets the weight of information detected by the laser radar to a value greater than that of information detected by the radio wave radar. Furthermore, in a driving scene in which a specific electromagnetic wave condition related to multiple reflection of electromagnetic waves is satisfied, the setting unit sets the weight of information detected by the camera to a value greater than that of information detected by the laser radar and the radio wave radar. In addition, when the driving scene satisfies specific real-scene reflection conditions regarding reflective objects that reflect the real outside scenery, the setting unit sets the weight of the detection information from the camera to be smaller than the weight of the detection information from the laser radar and the radio wave radar. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 7347343 Summary of the Invention [Problem to be solved by the invention]

[0005] In this type of device, various attempts have been made to further improve the recognition accuracy. The present disclosure has been made in consideration of the circumstances exemplified above. [Means for solving the problem]

[0006] In one aspect of the present disclosure, an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle includes: a sensor information acquisition unit (601) that acquires detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; an object detection unit (602) that generates an object detection result corresponding to the detection information acquired by the sensor information acquisition unit for each of the plurality of types of object detection sensors; an integration processing unit (604) that generates a detection performance curve for each of the plurality of types of object detection sensors, calculates reliability of the object detection results using the detection performance curve, and integrates the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; a recognition processing unit (605) that recognizes the object based on the integration result of the reliability by the integration processing unit; Equipped with The integration processing unit generates the detection performance curve based on feedback information of the vehicle's driving environment and / or the object recognition result by the recognition processing unit. In another aspect of the present disclosure, an object recognition method performed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle includes the following steps or processes: Acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; generating an object detection result corresponding to the detection information acquired by the sensor information acquisition unit for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; Calculating the reliability of the object detection result using the detection performance curve; integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; recognize the object based on the integration result of the reliability by the integration processing unit; The detection performance curve is generated based on feedback information of the vehicle's driving environment and / or the object recognition result by the recognition processing unit. In yet another aspect of the present disclosure, an object recognition program executed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle includes, as a process executed by the object recognition device, A process of acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; a process of generating an object detection result corresponding to the detection information acquired by the sensor information acquisition unit for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; A process of calculating the reliability of the object detection result using the detection performance curve; a process of integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; a process of recognizing the object based on the integration result of the reliability by the integration processing unit; Including, In the process of generating the detection performance curve, the detection performance curve is generated based on feedback information of the vehicle's running environment and / or the object recognition result by the recognition processing unit.

[0007] In addition, in each section of the application documents, each element may be assigned a reference symbol in parentheses. However, such reference symbols merely indicate an example of the correspondence between the element and the specific means described in the embodiments below. Therefore, the present disclosure is not limited in any way by the above-mentioned reference symbols. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a schematic diagram showing a state in which a vehicle equipped with an in-vehicle system including an object recognition device according to an embodiment of the present disclosure is traveling; [Figure 2]FIG. 2 is a block diagram showing a schematic configuration of the in-vehicle system shown in FIG. [Figure 3] 3 is a block diagram showing a schematic functional configuration of the object recognition device shown in FIG. 2. FIG. [Figure 4] FIG. 4 is a conceptual diagram showing an outline of an object recognition operation by the object recognition device shown in FIG. 3. [Figure 5] 4 is a flowchart showing an outline of integration processing realized by the integration processing function shown in FIG. 3. [Figure 6] 6 is a flowchart showing a specific example of the reliability calculation process shown in FIG. 5. [Figure 7] 7 is a graph showing an outline of a theoretical PR curve generated by the theoretical PR curve generation process shown in FIG. 6. [Figure 8] 7 is a flowchart showing a specific example of the theoretical PR curve generation process shown in FIG. 6. [Figure 9] 9 is a table used to calculate the coefficient Kem in the Kem calculation process shown in FIG. 8. [Figure 10] 9 is a flowchart showing a specific example of the Krm calculation process shown in FIG. 8. [Figure 11] 11 is a table used to calculate the coefficient Krm in the Krm calculation process shown in FIG. 10. DETAILED DESCRIPTION OF THE INVENTION

[0009] (Embodiment) Hereinafter, exemplary embodiments and specific examples of the present disclosure will be described with reference to the accompanying drawings as appropriate. Referring to FIG. 1, an in-vehicle system 1 is configured to be mounted on a vehicle V and to execute various operations in the vehicle V. Hereinafter, the vehicle V equipped with the in-vehicle system 1 will be referred to as the "host vehicle." Specifically, the in-vehicle system 1 is configured to recognize an object B around the host vehicle using an object detection sensor 2, and to execute driving assistance operations such as following a preceding vehicle and collision avoidance using a driving assistance device 3 according to the recognition result.

[0010] (In-vehicle system configuration) Referring to FIG. 2, the in-vehicle system 1 includes an object detection sensor 2, a driving assistance device 3, a vehicle state sensor 4, a locator 5, and an object recognition device 6. The object detection sensor 2, the driving assistance device 3, the vehicle state sensor 4, the locator 5, and the object recognition device 6 are connected to each other via an in-vehicle network so that they can exchange signals with each other. The in-vehicle network is configured to comply with a predetermined communication standard such as CAN (internationally registered trademark: international registration number 1048262A). CAN (internationally registered trademark) is an abbreviation for Controller Area Network. Note that the in-vehicle network may have a sub-network that complies with LIN, FlexRay, or the like, in addition to a main network that complies with CAN (internationally registered trademark). LIN is an abbreviation for Local Interconnect Network. The following describes the schematic configuration and function of each part of the in-vehicle system 1.

[0011] In this embodiment, the in-vehicle system 1 includes a plurality of object detection sensors 2, namely, a first sensor 21, a second sensor 22, a third sensor 23, and a fourth sensor 24. The first to fourth sensors 21 to 24 are configured to detect the object B using different detection principles. Specifically, for example, the first sensor 21 is configured as a so-called in-vehicle camera sensor equipped with an image sensor such as a CCD or CMOS. CCD stands for Charge Coupled Device. CMOS stands for Complementary Metal Oxide Semiconductor. For example, the second sensor 22 is configured as a so-called millimeter-wave radar sensor that detects the relative position and speed of the object B relative to the vehicle by emitting millimeter-wave radio waves and receiving waves reflected by the object B. For example, the third sensor 23 is configured as a so-called laser radar sensor that detects the distance and direction to the object B using laser light. For example, the fourth sensor 24 is configured as a so-called sonar that detects the distance and direction to the object B using ultrasonic waves.

[0012] The driving assistance device 3 has a configuration as a so-called driving assistance ECU that controls a driving force generation mechanism, a driving force transmission mechanism, a braking mechanism, etc. to perform driving assistance operations. ECU stands for Electronic Control Unit. Here, "driving assistance" refers to the in-vehicle system 1 continuously executing at least one of a longitudinal vehicle motion control subtask and a lateral vehicle motion control subtask. The longitudinal vehicle motion control subtask is starting, accelerating / decelerating, and stopping. The lateral vehicle motion control subtask is steering. In other words, in this specification, "driving assistance" includes "driving assistance" in the narrow sense, in which both the longitudinal vehicle motion control subtask and the lateral vehicle motion control subtask are not simultaneously executed, and "advanced driving assistance," in which both are simultaneously executed. In this way, the in-vehicle system 1 is configured to achieve Level 1 or Level 2 driving automation in the host vehicle as defined in the standard "SAE J3016" published by SAE International. SAE stands for Society of Automotive Engineers.

[0013] The vehicle state sensor 4 is configured to detect various quantities related to the driving state of the host vehicle. The "driving state" includes the driving operation state, driving behavior state, and driving environment state of the host vehicle. The "driving operation state" refers to the state of driving operation input of the host vehicle by the driver of the host vehicle or the driving assistance device 3, and includes, for example, the steering amount, throttle opening, brake operation amount, shift range, etc. The "driving behavior state" refers to the state related to the movement or behavior of the host vehicle, and includes, for example, the vehicle speed, acceleration, yaw rate, etc. The "driving environment state" refers to the environment around the host vehicle that is different from the state of existence of object B, which is the target of detection or sensing by the object detection sensor 2, and includes, for example, the illuminance, weather, road surface condition, etc. around the host vehicle.

[0014] Specifically, in this embodiment, the in-vehicle system 1 includes at least a vehicle speed sensor 41, a yaw rate sensor 42, a rain sensor 43, a moisture sensor 44, and an illuminance sensor 45 as vehicle condition sensors 4. The vehicle speed sensor 41 is configured to detect the vehicle speed of the host vehicle. The yaw rate sensor 42 is configured to detect the yaw rate of the host vehicle. The rain sensor 43 is a so-called raindrop sensor, and is configured to generate an output corresponding to the state of raindrops adhering to a predetermined area on the upper edge of the front windshield of the host vehicle. The moisture sensor 44 is configured to detect whether the road surface ahead of the host vehicle is dry or wet. The illuminance sensor 45 is configured to detect the illuminance around the host vehicle. These sensors were already well known at the time of filing this application, so detailed description thereof will be omitted in this specification.

[0015] The locator 5 is configured to measure the position of the vehicle. Specifically, the locator 5 has at least a satellite positioning function that measures the position of the vehicle by receiving a positioning signal transmitted from a positioning satellite. The locator 5 may also have an autonomous positioning function using an inertial sensor such as a gyro sensor or an acceleration sensor to improve the measurement accuracy of the vehicle's position in places where satellite radio waves are difficult to reach, such as inside a tunnel. Such an inertial sensor may be provided in the locator 5 or in the vehicle state sensor 4. For example, the "POSLV" positioning and orientation system for land vehicles manufactured by Applanix is ​​commercially available as a locator 5 equipped with an inertial sensor. The locator 5 is also configured to be able to acquire information such as the road type and curve curvature at the current traveling position of the vehicle based on the positioning result of the vehicle and map information.

[0016] (object recognition device) The object recognition device 6 is configured as an on-board computer, i.e., an object recognition ECU, capable of performing so-called sensor fusion processing, which integrates detection results from multiple types of object detection sensors 2. Specifically, the object recognition device 6 includes a processor 61, which is a CPU or MPU, and a storage medium 62 communicably connected to the processor 61. The object recognition device 6 is configured so that the processor 61 reads and executes a computer program from the storage medium 62, thereby realizing a predetermined function for recognizing an object B around the vehicle. The storage medium 62 includes at least a ROM or a nonvolatile rewritable memory among various non-transient tangible storage media such as a ROM or a nonvolatile rewritable memory. The nonvolatile rewritable memory is a storage device that allows information to be rewritten while the power is on but retains information in an unrewritable manner while the power is off, such as a flash memory. The storage medium 62 stores the computer program as well as various data required to execute the program, such as initial values, maps, and look-up tables.

[0017] 3, the object recognition device 6 has, as functional configurations realized on an on-board microcomputer by executing a computer program, a sensor information acquisition function 601, an object detection function 602, a driving environment determination function 603, an integrated processing function 604, and an object recognition processing function 605. Details of these functional configurations will be explained below in order.

[0018] The sensor information acquisition function 601 and the object detection function 602 are provided corresponding to each of the multiple object detection sensors 2. The sensor information acquisition function 601, which corresponds to the sensor information acquisition unit in the present disclosure, acquires, i.e., receives, detection information from the corresponding object detection sensor 2. The object detection function 602, which corresponds to the object detection unit in the present disclosure, generates an object detection result corresponding to the detection information acquired by the sensor information acquisition function 601. The driving environment determination function 603 determines the driving environment of the host vehicle based on output information acquired from the object detection sensor 2 or the vehicle state sensor 4. Details of the driving environment determination will be described later. The integration processing function 604, which corresponds to the integration processing unit in the present disclosure, integrates the detection information from each of the multiple object detection sensors 2. The object recognition processing function 605, which corresponds to the recognition processing unit in the present disclosure, performs object recognition based on the integration result by the integration processing function 604. In addition, the object recognition processing function 605 outputs feedback information of the object recognition result to the integration processing function 604.

[0019] (Sensor fusion processing overview) FIG. 4 shows an overview of the sensor fusion process used for object recognition in this embodiment, i.e., the processing content of the integration processing function 604 shown in FIG. 3. This sensor fusion process uses DBF, a rectangle-level integration method. DBF stands for Dynamic Belief Fusion. DBF models three hypotheses, "target," "non-target," and "intermediate state," by calculating confidence from a precision-recall curve based on the detection score s for each of multiple detection results, thereby achieving more accurate fusion. The detection score s is a value indicating the likelihood of the object being recognized, i.e., the likelihood of it being a vehicle or a pedestrian. For details about DBF, see the following paper: "DBF: Dynamic Belief Fusion for Combining Multiple Object Detectors," IEEE Transactions on Pattern Analysis and Machine Intelligence (Volume: 43, Issue: 5), Page(s): 1499-1514.

[0020] As a specific example, Fig. 4 shows a case where an automobile is recognized as another vehicle from sensor detection information. In Fig. 4, first sensor information 711 is detection information from the first sensor 21. A first detection rectangle 712 is a bounding box provided in the first sensor information 711. As object detection result information in the object detection function 602 shown in Fig. 3, as shown in Fig. 4, the detected object within the first detection rectangle 712 is an automobile as a "target," and a detection score s for being an automobile is output.

[0021] The first reliability calculation function 713 calculates the reliability, i.e., basic probability, for three hypotheses, "target," "non-target," and "intermediate state," for the object detection result information based on the detection information of the first sensor 21. The reliability calculation results for each of the three hypotheses by the first reliability calculation function 713 are shown as first reliability allocation information 714. The first reliability allocation information 714 includes basic probability assignments p(T), p(NT), and p(I), for the detected object within the first detection rectangle 712. p(T) indicates the basic probability that the detected object within the first detection rectangle 712 is a "target," i.e., a vehicle. p(NT) indicates the basic probability that the detected object within the first detection rectangle 712 is a "non-target," i.e., a non-vehicle. p(I) indicates the basic probability that the detected object within the first detection rectangle 712 is in an "intermediate state," i.e., it is unknown whether the object is a vehicle or not. The concepts of "basic probability" and "basic probability allocation" are based on the Dempster-Shafer theory and were already well known at the time of filing this application, so detailed explanations (e.g., mathematical formulas) other than those described below will be omitted in this specification.

[0022] Similarly, second sensor information 721 is detection information from the second sensor 22. A second detection rectangle 722 is a bounding box provided in the second sensor information 721. A second reliability calculation function 723 calculates the reliability of three hypotheses, "target," "non-target," and "intermediate state," for the object detection result information based on the detection information from the second sensor 22. The reliability calculation results for each of the three hypotheses by the second reliability calculation function 723 are shown as second reliability allocation information 724. The second reliability allocation information 724 includes p(T), p(NT), and p(I), which are basic probability allocations for the detected object within the second detection rectangle 722. Similar processing is performed on the detection information from the third sensor 23 and the fourth sensor 24, and the corresponding basic probability allocations are output.

[0023] Basic probability allocation will be described using the diagram shown in the lower right of FIG. 4 in which the third reliability calculation function 733 outputs third reliability allocation information 734 based on the detection information of the third sensor 23. First, the third reliability calculation function 733 generates a detection performance curve for the corresponding object detection sensor 2, i.e., the third sensor 23. In this embodiment, the detection performance curve is a precision-recall curve. The precision-recall curve will be abbreviated as a "PR curve" hereinafter. In the "PR curve," "P" indicates precision (i.e., precision) and "R" indicates recall (i.e., recall).

[0024] Specifically, the third reliability calculation function 733 generates PR curve information 751. The PR curve information 751 includes a theoretical PR curve PRi, which is a PR curve that estimates the theoretical performance of the third sensor 23, and an actual PR curve PRr, which is a PR curve that represents the actual performance of the third sensor 23. The theoretical PR curve PRi is calculated by the following equation (1): n " The above formula (1) is quoted from the above paper.

number

[0025] In the PR curve information 751, the value on the performance PR curve PRr at the recall r(s) corresponding to the detection score s (i.e., the difference between the baseline at precision=0 at the recall r(s) and the performance PR curve PRr) is p(T). p(NT) is the value shown in the following equation (2). That is, p(NT) is the difference between the theoretical PR curve PRi at the recall r(s) and the horizontal line at precision=1. p(I) is the value shown in the following equation (3). That is, p(I) is the difference between the performance PR curve PRr and the theoretical PR curve PRi at the recall r(s).

number

number

[0026] Allocation function information 752 illustrated between the PR curve information 751 and the third reliability allocation information 734 shows p(T), p(NT), and p(I) as a function of the detection score s. The third reliability allocation information 734 indicates the basic probability allocation p(T), p(NT), and p(I) at the detection score s. The calculated basic probability allocations, namely, the first reliability allocation information 714, the second reliability allocation information 724, the third reliability allocation information 734, etc., are combined, i.e., integrated, by a combination rule 753. The combination rule 753 may use, for example, the Dempster combination rule.

[0027] In this embodiment, the exponent parameter n in the theoretical PR curve PRi shown in the above equation (1) is determined or adjusted in accordance with feedback information on object recognition results and the driving scene, thereby further improving recognition accuracy. That is, the integration processing function 604 generates the theoretical PR curve PRi as a detection performance curve based on the driving environment of the vehicle and / or feedback information on object recognition results by the object recognition processing function 605. The integration processing function 604 also uses the generated theoretical PR curve PRi to calculate the reliability of the object detection results, i.e., basic probability allocation, for each of the multiple types of object detection sensors 2, and integrates these. The object recognition processing function 605 then performs object recognition based on the integration result.

[0028] (Example of operation) Below, with reference to Figures 1 to 4 as well as Figure 5 onwards, an overview of the operation of the object recognition device 6 according to this embodiment will be described, along with the effects achieved by the object recognition method and object recognition program executed thereby. Note that in the flowcharts shown in Figure 5 etc., "S" is an abbreviation for "step". Furthermore, the object recognition device 6 according to this embodiment and the object recognition method and object recognition program executed thereby may hereinafter be collectively referred to as "this embodiment".

[0029] Fig. 5 shows an overview of the integration process by the integration processing function 604 shown in Fig. 3. First, in step 101, the processor 61 in the object recognition device 6 calculates the reliability of the object detection results, i.e., basic probability allocation, for each of the multiple types of object detection sensors 2. Next, in step 102, the processor 61 integrates the calculated basic probability allocations.

[0030] FIG. 6 shows the specific contents of the reliability calculation process in step 101. First, in step 201, the processor 61 determines whether or not calculation of the reliability of the object detection results, i.e., basic probability allocation, has been completed for all of the multiple types of object detection sensors 2, i.e., the first sensor 21 to the fourth sensor 24. The determination result in step 201 remains "NO" until calculation is completed for all of the multiple types of object detection sensors 2, and the processor 61 executes the processes of steps 202 and 203, and then returns the process to step 201. When calculation is completed for all of the multiple types of object detection sensors 2, the determination result in step 201 remains "YES," and the processor 61 temporarily terminates the process of the flowchart shown in FIG. 6 and proceeds to step 102 shown in FIG. 5. In step 202, the processor 61 generates a theoretical PR curve PRi. In step 203, the processor 61 calculates basic probability allocations p(T), p(NT), and p(I) using the theoretical PR curve PRi generated in step 202.

[0031] The generation of the theoretical PR curve PRi in step 202 will be described with reference to Fig. 7 and Fig. 8. Fig. 7 shows the theoretical PR curve PRi corresponding to various values ​​of the exponential parameter n. As shown in Fig. 7, when n = 1, the curve is a straight line sloping downward to the right, when n < 1, the curve is convex toward the lower left, when n > 1, the curve is convex toward the upper right, and when n = ∞, the precision rate is 1 when the recall rate r < 1, and when the recall rate r = 1, the precision rate is a rectangular wave that drops from 1 to 0.

[0032] 8 shows the specific contents of the theoretical PR curve generation process in step 202. First, in step 301, the processor 61 sets the exponent parameter n to the initial value n init Set the initial value n init is a predetermined value according to the type of object detection sensor 2, and can be obtained by, for example, optimization experiments or computer simulations. Next, in step 302, the processor 61 calculates the initial value n init First coefficient Ke to multiply by m Subsequently, in step 303, the processor 61 calculates the initial value n init The second coefficient Kr to multiply by m Then, in step 304, processor 61 calculates the initial value n init The first coefficient Ke m and the second coefficient Kr m The exponent parameter n is determined by multiplying

[0033] First coefficient Ke m is a coefficient according to the driving scene or driving environment of the vehicle, and is the weather coefficient K wx , time zone coefficient K tm , solar position coefficient K sp , road type coefficient K rt , road surface condition coefficient K rc , etc. If it is determined that the current driving scene or driving environment of the host vehicle has a negative effect on the detection and recognition of object B by the object detection sensor 2, the risk of false detection increases. Therefore, in this case, it is preferable to reduce the value of the exponent parameter n to increase the reliability of non-targets. Therefore, the first coefficient Ke m These coefficients constituting the coefficients are numerical values ​​exceeding 0 and equal to or less than 1, and are set so that the worse the conditions, the smaller the value. The driving scene or driving environment can be determined based on information detected by the object detection sensor 2 (for example, images of the sky or raindrops taken by a camera), output information from the vehicle condition sensor 4, output information from the locator 5, road traffic information or weather information acquired from an external server, etc. Details of these representative coefficients will be explained below in order.

[0034] Weather coefficient K wx In the case of clear weather, the first sensor 21 is K 1_fw , and K for the second sensor 22 2_fw , and for the third sensor 23, K 3_fw , and for the fourth sensor 24, K 4_fw In the case of cloudy weather, the first sensor 21 is set to K 1_cw , and K for the second sensor 22 2_cw , and for the third sensor 23, K 3_cw , and for the fourth sensor 24, K 4_cw In the case of rain, the first sensor 21 is set to K 1_rw , and K for the second sensor 22 2_rw , and for the third sensor 23, K 3_rw , and for the fourth sensor 24, K 4_rw These weather coefficients K wx The value of is set using optimization experiments, computer simulations, etc., based on the strengths and weaknesses of each of the first sensor 21 to the fourth sensor 24. The determination of whether the weather is sunny or cloudy can be made based on, for example, camera images, the output of the rain sensor 43, or weather information obtained from an external server.

[0035] Time zone coefficient K tm In the daytime, the first sensor 21 is K 1_dt , and K for the second sensor 22 2_dt , and for the third sensor 23, K 3_dt , and for the fourth sensor 24, K 4_dt At night, the first sensor 21 is set to K 1_nt , and K for the second sensor 22 2_nt , and for the third sensor 23, K 3_nt , and for the fourth sensor 24, K 4_nt These time zone coefficients K tmThe value of is set using optimization experiments, computer simulations, etc., based on the strengths and weaknesses of each of the first sensor 21 to the fourth sensor 24. The time period can be determined using, for example, output information from locator 5 or output information from a clock (not shown) mounted on the vehicle that can output date information and time information.

[0036] Solar position coefficient K sp In the case of direct sunlight, the first sensor 21 is K 1_fl , and K for the second sensor 22 2_fl , and for the third sensor 23, K 3_fl , and for the fourth sensor 24, K 4_fl In the case of backlight, the first sensor 21 is set to K 1_bl , and K for the second sensor 22 2_bl , and for the third sensor 23, K 3_bl , and for the fourth sensor 24, K 4_bl These solar position coefficients K sp The value of is set using optimization experiments, computer simulations, etc., based on the degree of strength and weakness of each of the first sensor 21 to the fourth sensor 24. The determination of whether the light is from the sun or from the back can be made based on, for example, the orientation of the host vehicle obtained from camera images, or output information from the vehicle state sensor 4 and / or locator 5, and the altitude and direction of the sun calculated from the date and time zone.

[0037] Road type coefficient K rt When the vehicle is traveling on a public road, the first sensor 21 is K 1_gr , and K for the second sensor 22 2_gr , and for the third sensor 23, K 3_gr , and for the fourth sensor 24, K 4_gr When the vehicle is traveling on a highway, the first sensor 21 is set to K 1_ar , and K for the second sensor 22 2_ar , and for the third sensor 23, K 3_ar , and for the fourth sensor 24, K 4_ar When traveling in a tunnel, the first sensor 21 is set to K 1_tl, and K for the second sensor 22 2_tl , and for the third sensor 23, K 3_tl , and for the fourth sensor 24, K 4_tl These road type coefficients K rt The value of is set using optimization experiments, computer simulations, etc., based on the degree of strength or weakness of each of the first sensor 21 to the fourth sensor 24. The road type can be determined based on, for example, the output information from the locator 5 and map information.

[0038] Road condition coefficient K rc In the case of a dry road surface, the first sensor 21 is K 1_dy , and K for the second sensor 22 2_dy , and for the third sensor 23, K 3_dy , and for the fourth sensor 24, K 4_dy In the case of a wet road surface, the first sensor 21 is set to K 1_wt , and K for the second sensor 22 2_wt , and for the third sensor 23, K 3_wt , and for the fourth sensor 24, K 4_wt These road surface condition coefficients K rc The value of is set using optimization experiments, computer simulations, etc., based on the degree of strength or weakness of each of the first sensor 21 to the fourth sensor 24. The road surface condition can be determined based on, for example, camera images or output information from the moisture sensor 44.

[0039] Second coefficient Kr m is a coefficient corresponding to feedback information of the object recognition result by the object recognition processing function 605, i.e., the object recognition result from one time point before. "One time point" is one period of the object recognition cycle (for example, 10 msec). A high similarity between the predicted value at the current time point (i.e., this time) based on the object recognition result from one time point before (i.e., the previous time) and the detected value at the current time point indicates that the target object has been tracked with high accuracy. Therefore, in this case, it is preferable to increase the value of the exponent parameter n to reduce the reliability of non-targets. Therefore, the second coefficient Kr mEach coefficient constituting the above equation (described later) is a numerical value exceeding 0 and equal to or less than 1, and is set so that the value increases as the target object is tracked more accurately.

[0040] FIG. 10 shows the second coefficient Kr m The specific content of the calculation process will be described below. First, in step 401, the processor 61 calculates the exponent parameter n, i.e., the second coefficient Kr m Until calculations for all of the plurality of types of object detection sensors 2 are completed, the determination result in step 401 will be “NO”, and processor 61 will proceed to step 402.

[0041] In step 402, the processor 61 determines whether the detection result corresponding to the current detection score s is within the predicted range. That is, the processor 61 determines whether the bounding box in the current detection information is within the predicted range calculated based on the bounding box in the previous detection information. The predicted range can be calculated by taking into account the relative movement direction and relative movement speed between the host vehicle and the target object, etc., with respect to the previous detection information. Specifically, first, a predicted bounding box is generated using the position, width, depth, orientation, speed, and angular velocity of the previous detection information. Next, a margin is added to this predicted bounding box in the width and depth directions to generate a bounding box that is one size larger, and this is set as the predicted range. If the determination result in step 402 is "NO," the processor 61 returns the process to step 401. On the other hand, if the determination result in step 402 is "YES," the processor 61 executes the processes of steps 403 and 404 and then returns the process to step 401.

[0042] In step 403, processor 61 extracts a predicted value that is closest to the detection result corresponding to the current detection score s. That is, processor 61 sets a bounding box of the same size as the bounding box in the previous detection information so that it is closest to the bounding box in the current detection information within the prediction range, and sets this as the predicted value. In other words, processor 61 sets as the predicted value the closest predicted bounding box in which the current target detection result overlaps the prediction range and is closest in distance to the predicted bounding box. Parameters such as position and size related to the bounding box in the current detection information will be referred to as "measured values" hereinafter.

[0043] In step 404, the processor 61 calculates the second coefficient Kr based on the predicted value and the measured value. m Calculate the second coefficient Kr m is the position residual coefficient K pos , size residual coefficient K siz , orientation residual coefficient K yaw , velocity residual coefficient K vel , tracking count coefficient K age , attribute coefficient K class , etc. These representative coefficients will be explained below in order.

[0044] Position residual coefficient K pos is a coefficient related to the position residual, i.e., the difference between the predicted value and the measured value of the reference point of the bounding box (e.g., one of the four corner points or the center point). pos1 If it is less than the position residual coefficient K pos =K pos1 When the position residual is medium, that is, the first position residual threshold Th pos1 and the second position residual threshold Th pos2 If the position residual coefficient K is between pos =K pos2 If the position residual is large, that is, if the second position residual threshold Th pos2 If it is greater than or equal to the position residual coefficient K pos =K pos3 Let 1>Kpos1 >K pos2 >K pos3 >0.

[0045] Size residual coefficient K siz is a coefficient related to the size residual, i.e., the difference between the predicted and measured bounding box size. When the size residual is small, i.e., the first size residual threshold Th siz1 If it is less than the size residual coefficient K siz =K siz1 When the size residual is medium, that is, the first size residual threshold Th siz1 and the second size residual threshold Th siz2 If the size residual coefficient K is between siz =K siz2 The size residual is large, i.e., the second size residual threshold Th siz2 If it is greater than or equal to the size residual coefficient K siz =K siz3 Let 1>K siz1 >K siz2 >K siz3 >0.

[0046] Orientation residual coefficient K yaw is a coefficient related to the orientation residual, i.e., the difference between the predicted and measured orientation of the bounding box. yaw1 If it is less than the orientation residual coefficient K yaw =K yaw1 If the orientation residual is medium, that is, the first orientation residual threshold Th yaw1 and the second directional residual threshold Th yaw2 If the relationship is between yaw =K yaw2 The orientation residual is large, that is, the second orientation residual threshold Th yaw2 If it is greater than or equal to the yaw =K yaw3 Let 1>K yaw1 >K yaw2 >K yaw3 >0.

[0047] Velocity residual coefficient K velis a coefficient related to the velocity residual, i.e., the difference between the predicted and measured values ​​of the movement velocity of the bounding box. vel1 If it is less than the velocity residual coefficient K vel =K vel1 When the velocity residual is medium, that is, the first velocity residual threshold Th vel1 and the second velocity residual threshold Th vel2 If the velocity residual coefficient K is between vel =K vel2 The velocity residual is large, that is, the second velocity residual threshold Th vel2 If it is greater than or equal to the velocity residual coefficient K vel =K vel3 If there is no velocity residual, i.e., it is 0, the velocity residual coefficient K vel = 1. 1>K vel1 >K vel2 >K vel3 >0.

[0048] Tracking count coefficient K age is a coefficient related to the number of times the same object is tracked. age1 If it is less than the tracking count coefficient K age =K age1 The number of tracking attempts is medium, i.e., the first tracking attempt threshold Th age1 and the second tracking count threshold Th age2 If the tracking count coefficient K age =K age2 The number of tracking times is large, that is, the second tracking time threshold Th age2 If it is equal to or greater than the tracking count coefficient K age =K age3 Let 1>K age3 >K age2 >K age1 >0.

[0049] Attribute Coefficient K class is a coefficient related to the match / mismatch of attributes in detected objects. "Attributes" are, for example, cars, pedestrians, etc. In the case of attribute match, the attribute coefficient K class =K class1 In the case of attribute mismatch, the attribute coefficient K class=K class2 Let 1>K class1 >K class2 >0.

[0050] As described above, this embodiment generates a theoretical PR curve PRi, which is a PR curve that estimates the theoretical performance of the object detection sensor 2, and an actual PR curve PRr, which is a PR curve that represents the actual performance of the object detection sensor 2. Furthermore, in generating the theoretical PR curve PRi for each of the multiple types of object detection sensors 2, this embodiment determines the exponent parameter n based on the current driving scene or driving environment of the vehicle and feedback information on the previous recognition result, i.e., one time before. This makes it possible to generate an appropriate detection performance curve for each of the multiple types of object detection sensors 2 that is appropriate for the current operating situation. Therefore, this embodiment makes it possible to further improve the recognition accuracy of object recognition based on sensor fusion processing that applies the Dempster-Shafer theory or DBF.

[0051] (Variation) The present disclosure is not limited to the above-described embodiments and specific examples. Therefore, the above-described embodiments and the like can be modified as appropriate. Representative modifications will be described below. In the following description of the modifications, differences from the above-described embodiments and the like will be mainly described. Furthermore, the same reference numerals are used for parts that are identical or equivalent to each other in the above-described embodiments and the following modifications. Therefore, in the following description of the modifications, the explanations in the above-described embodiments and the like can be used as appropriate for components that have the same reference numerals as the above-described embodiments and the like, unless there is a technical contradiction or special additional explanation.

[0052] The present disclosure is not limited to the specific applications and device configurations shown in the above embodiments. That is, for example, the host vehicle may be a so-called automobile or a motorcycle. There are no particular limitations on the type of automobile or motorcycle. Furthermore, "object B" is not limited to targets for following a preceding vehicle or collision avoidance, i.e., other vehicles or obstacles, but may be, for example, road signs, traffic lights, road markings, etc. That is, "object B" may also be referred to as "target." Therefore, the object recognition device 6 according to the present disclosure may also be referred to as a "target recognition device."

[0053] The in-vehicle system 1 may be configured to achieve "autonomous driving" of level 3 or higher as defined in "SAE J3016." In this case, the driving assistance device 3 has a configuration as a so-called autonomous driving ECU. The types of object detection sensors 2 are not limited to the four types exemplified in the above embodiment, but may be two, three, five or more. Furthermore, the present disclosure may be suitably applied to fusion processing of multiple object detection sensors 2 having the same detection principle but different effective detection ranges, such as long-range radar, medium-range radar, and short-range radar. In this sense, multiple object detection sensors 2 having the same detection principle but different effective detection ranges or operating frequency bands can be evaluated as "multiple types of object detection sensors 2."

[0054] All or part of the object recognition device 6 may be configured with a digital circuit, such as an ASIC or FPGA, configured to be able to realize the above-described functions or operations. ASIC stands for Application Specific Integrated Circuit. FPGA stands for Field Programmable Gate Array. In other words, the object recognition device 6 may include both an on-board microcomputer and a digital circuit.

[0055] The program according to the present disclosure, which enables the execution of various operations, procedures, or processes described in the above embodiments, can be downloaded or upgraded via V2X communication. V2X stands for Vehicle to X. Alternatively, the program can be downloaded or upgraded via a terminal device installed in a vehicle manufacturing plant, a repair shop, a dealer, or the like. The program can be stored on a memory card, an optical disk, a magnetic disk, or the like.

[0056] In this way, each of the above functional configurations and processes may be realized by a special-purpose computer provided by configuring a processor 61 and a storage medium 62 programmed to execute one or more functions embodied in a computer program. Alternatively, each of the above functional configurations and processes may be realized by a special-purpose computer provided by configuring a processor 61 with one or more dedicated hardware logic circuits. Alternatively, each of the above functional configurations and processes may be realized by one or more special-purpose computers configured by combining one or more processors 61 programmed to execute one or more functions, one or more storage media 62, and one or more other processors 61 configured with one or more hardware logic circuits. Furthermore, a computer program may be stored in a computer-readable, non-transitory storage medium as instructions to be executed by a computer. In other words, each of the above functional configurations and processes may be expressed as a computer program including procedures for implementing the program, or as a non-transitory storage medium storing the program.

[0057] The present disclosure is not limited to the specific functions and operational modes shown in the above embodiment. That is, for example, other well-known combination rules such as the Yager combination rule can be used as the combination rule 753 instead of the Dempster combination rule. Also, the flowcharts shown in the figures can be modified as appropriate. Specifically, for example, the first coefficient Ke m and the second coefficient Krm One of these may be omitted. That is, the object recognition device 6 may determine the exponent parameter n using only either the current driving scene or driving environment of the host vehicle or the feedback information of the previous recognition result, i.e., one time instant before. Also, in step 403, the predicted value, i.e., the closest bounding box may be set to the predicted bounding box.

[0058] Similar expressions such as "acquire," "calculate," "estimate," "detect," and "sensing" may be substituted for each other as appropriate within the scope of technical inconsistency. Furthermore, "exceeding the threshold" and "above the threshold" may be substituted for each other as appropriate within the scope of technical inconsistency. The same applies to "below the threshold" and "below the threshold."

[0059] It goes without saying that the elements constituting the above-described embodiments are not necessarily essential unless expressly stated as essential or clearly considered essential in principle. Furthermore, when numerical values ​​such as the number, value, amount, and range of components are mentioned, the present disclosure is not limited to those specific numbers unless expressly stated as essential or clearly limited to a specific number in principle. Similarly, when the shape, direction, positional relationship, etc. of components are mentioned, the present disclosure is not limited to those shapes, directions, positional relationships, etc. unless expressly stated as essential or clearly limited to a specific shape, direction, positional relationship, etc. in principle.

[0060] The modified examples are not limited to the above examples. For example, all or part of one of the multiple specific examples may be combined with all or part of another of the multiple specific examples, provided that there is no technical inconsistency. There is no particular limit to the number of combinations. Similarly, all or part of one of the multiple modified examples may be combined with all or part of another of the multiple modified examples, provided that there is no technical inconsistency. Furthermore, all or part of the above specific example and all or part of the above modified examples may be combined with each other, provided that there is no technical inconsistency.

[0061] (Disclosure perspective) As is clear from the above description of the embodiments and modifications, this specification discloses at least the following matters.

[0062] [Point 1-1] An object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, a sensor information acquisition unit (601) that acquires detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; an object detection unit (602) that generates an object detection result corresponding to the detection information acquired by the sensor information acquisition unit for each of the plurality of types of object detection sensors; an integration processing unit (604) that generates a detection performance curve for each of the plurality of types of object detection sensors, calculates reliability of the object detection results using the detection performance curve, and integrates the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; a recognition processing unit (605) that recognizes the object based on the integration result of the reliability by the integration processing unit; Equipped with The integrated processing unit generates the detection performance curve based on feedback information of the vehicle's running environment and / or the object recognition result by the recognition processing unit. Object recognition device. [Point 1-2] The integration processing unit generates, as the detection performance curves, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor. An object recognition device according to aspect 1-1. [Points 1-3] The integration processing unit determines the theoretical PR curve for each of the plurality of types of object detection sensors based on the traveling environment. An object recognition device according to aspect 1-2. [Points 1-4] the integration processing unit determines the theoretical PR curve for each of the plurality of types of object detection sensors based on the feedback information. An object recognition device according to aspect 1-2 or aspect 1-3. [Points 1-5] The integration processing unit is n " to determine the theoretical PR curve for each of the plurality of types of object detection sensors. An object recognition device according to aspect 1-3 or aspect 1-4.

[0063] [Point 2-1] An object recognition method performed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, comprising: Acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; generating an object detection result corresponding to the acquired detection information for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; Calculating the reliability of the object detection result using the detection performance curve; integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; Recognizing the object based on the integration result of the confidence levels; The detection performance curve is generated based on feedback information of the vehicle's driving environment and / or object recognition results. Object recognition method. [Point 2-2] As the detection performance curves, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor are generated. The object recognition method according to aspect 2-1. [Point 2-3] determining the theoretical PR curve for each of the plurality of types of object detection sensors based on the traveling environment; The object recognition method according to aspect 2-2. [Point 2-4] determining the theoretical PR curve for each of the plurality of types of object detection sensors based on the feedback information; The object recognition method according to aspect 2-2 or aspect 2-3. [Point 2-5] "1-r n " to determine the theoretical PR curve for each of the plurality of types of object detection sensors. The object recognition method according to aspect 2-3 or aspect 2-4.

[0064] [Point 3-1] An object recognition program executed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, The process executed by the object recognition device is A process of acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; generating an object detection result corresponding to the acquired detection information for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; A process of calculating the reliability of the object detection result using the detection performance curve; a process of integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; A process of recognizing the object based on the integration result of the reliability; Including, In the process of generating the detection performance curve, the detection performance curve is generated based on feedback information of the vehicle's running environment and / or object recognition results. Object recognition program. [Point 3-2] In the process of generating the detection performance curve, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor are generated as the detection performance curve. An object recognition device according to aspect 3-1. [Point 3-3] In the process of generating the detection performance curve, the theoretical PR curve for each of the plurality of types of object detection sensors is generated based on the traveling environment. An object recognition device according to aspect 3-2. [Point 3-4] In the process of generating the detection performance curve, the theoretical PR curve for each of the plurality of types of object detection sensors is generated based on the feedback information. An object recognition device according to aspect 3-2 or aspect 3-3. [Point 3-5] In the process of generating the detection performance curve, n " to generate the theoretical PR curve for each of the plurality of types of object detection sensors. An object recognition device according to aspect 3-3 or aspect 3-4. [Explanation of symbols]

[0065] 2. Object detection sensor 6 Object recognition device 601 Sensor information acquisition function 602 Object detection function 604 Integrated Processing Function 605 Object Recognition Processing Function B Object PRi Theoretical PR curve PRr Ability PR curve V vehicle

Claims

1. An object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, a sensor information acquisition unit (601) that acquires detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; an object detection unit (602) that generates an object detection result corresponding to the detection information acquired by the sensor information acquisition unit for each of the plurality of types of object detection sensors; an integration processing unit (604) that generates a detection performance curve for each of the plurality of types of object detection sensors, calculates a reliability of the object detection result using the detection performance curve, and integrates the calculated reliability of the object detection result for each of the plurality of types of object detection sensors; a recognition processing unit (605) that recognizes the object based on the integration result of the reliability by the integration processing unit; Equipped with The integration processing unit generates the detection performance curve based on feedback information of a running environment of the vehicle and / or an object recognition result by the recognition processing unit. Object recognition device.

2. The integration processing unit generates, as the detection performance curves, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor. The object recognition device according to claim 1 .

3. the integration processing unit determines the theoretical PR curve for each of the plurality of types of object detection sensors based on the traveling environment. The object recognition device according to claim 2 .

4. the integration processing unit determines the theoretical PR curve for each of the plurality of types of object detection sensors based on the feedback information. The object recognition device according to claim 2 or 3.

5. The integration processing unit is ｎ " to determine the theoretical PR curve for each of the plurality of types of object detection sensors. The object recognition device according to claim 4 .

6. An object recognition method performed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, comprising: Acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; generating an object detection result corresponding to the acquired detection information for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; Calculating the reliability of the object detection result using the detection performance curve; integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; Recognizing the object based on the integration result of the confidence levels; The detection performance curve is generated based on feedback information of the vehicle's driving environment and / or object recognition results. Object recognition method.

7. As the detection performance curves, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor are generated. The object recognition method according to claim 6.

8. determining the theoretical PR curve for each of the plurality of types of object detection sensors based on the traveling environment; The object recognition method according to claim 7 .

9. determining the theoretical PR curve for each of the plurality of types of object detection sensors based on the feedback information; The object recognition method according to claim 7 or 8.

10. "1-r ｎ " to determine the theoretical PR curve for each of the plurality of types of object detection sensors. The object recognition method according to claim 9 .

11. An object recognition program executed by an object recognition device (6) configured to be mounted on a vehicle (V) and recognize an object (B) around the vehicle, The process executed by the object recognition device is A process of acquiring detection information from each of a plurality of types of object detection sensors (2) mounted on the vehicle; generating an object detection result corresponding to the acquired detection information for each of the plurality of types of object detection sensors; generating a detection performance curve for each of the plurality of types of object detection sensors; A process of calculating the reliability of the object detection result using the detection performance curve; a process of integrating the calculated reliability of the object detection results for each of the plurality of types of object detection sensors; A process of recognizing the object based on the integration result of the reliability; Including, In the process of generating the detection performance curve, the detection performance curve is generated based on feedback information of a running environment of the vehicle and / or an object recognition result. Object recognition program.

12. In the process of generating the detection performance curve, a theoretical PR curve (PRi) which is a precision-recall curve that estimates the theoretical performance of the object detection sensor, and a performance PR curve (PRr) which is a precision-recall curve that represents the actual performance of the object detection sensor are generated as the detection performance curve. The object recognition device according to claim 11.

13. In the process of generating the detection performance curve, the theoretical PR curve for each of the plurality of types of object detection sensors is generated based on the traveling environment. The object recognition device according to claim 12.

14. In the process of generating the detection performance curve, the theoretical PR curve for each of the plurality of types of object detection sensors is generated based on the feedback information. The object recognition device according to claim 12 or 13.

15. In the process of generating the detection performance curve, ｎ " to generate the theoretical PR curve for each of the plurality of types of object detection sensors. The object recognition device according to claim 14.

Images