Target detection method and device, vehicle control unit and vehicle

Abstract

The embodiment of the invention provides a target detection method and device, a vehicle control unit and a vehicle. First target detection result sets corresponding to a first fusion target at multiple moments are obtained; wherein the first fusion target is obtained by fusing all first target detection results in the first target detection result set; based on the first target detection result set, determining a target associated cost value of a sensor set formed by sensors corresponding to the first target detection result; determining a target sensor from the sensor set based on the target associated cost values corresponding to the sensor set at the plurality of moments; and independently determining the first target through the first target detection result of the target sensor. According to the method, the target fusion result corresponding to the sensor generating the fusion target can be detected, and the sensor which is not suitable for continuously generating the fusion target can be found in time, so that the target detection results of the sensors are stripped to independently carry out target detection, and the accuracy of target detection can be improved.

Description

Technical Field

[0001] This application relates to the field of target detection technology, and in particular to a target detection method, device, vehicle controller, and vehicle. Background Technology

[0002] With the continuous development of technology, autonomous driving and driver assistance technologies are being used more and more widely in modern vehicles. In order to improve the vehicle's ability to recognize and navigate in complex environments, the integration of target detection technology has received increasing attention.

[0003] In related technologies, methods such as ID matching are typically used to match target detection results corresponding to the same entity from various sensors, and then these matching target detection results are fused to determine the fused target.

[0004] However, during continuous operation of the sensor, the correspondence between the target detection results and the real object, as well as the accuracy of the target detection results, will change, resulting in poor accuracy of the fused target obtained based on the above matching relationship. Summary of the Invention

[0005] In view of this, this application aims to provide a target detection method, apparatus, vehicle controller and vehicle to solve the problem of poor target detection accuracy in the prior art of multi-sensor fusion.

[0006] To achieve the above objectives, the technical solution of this application is implemented as follows:

[0007] Firstly, this application provides a target detection method, the method comprising:

[0008] Obtain a set of first target detection results corresponding to the first fusion target at multiple time points; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set;

[0009] Based on the first target detection result set, determine the target association value of the sensor set consisting of the sensors corresponding to the first target detection results respectively;

[0010] Based on the target association value corresponding to the sensor set at the multiple times, the target sensor is determined from the sensor set;

[0011] The first target is independently determined based on the first target detection result of the target sensor.

[0012] Optionally, determining the target association value of the sensor set consisting of the sensors corresponding to the first target detection results based on the first target detection result set includes:

[0013] Based on the first target detection result, determine the first association value between the first sensor and the second sensor other than the first sensor in the sensor set;

[0014] Based on the first association generation value and the number of sensors in the sensor set, the target association generation value corresponding to the sensor set is determined.

[0015] Optionally, determining the target sensor from the sensor set based on the target association value corresponding to the sensor set at the plurality of times includes:

[0016] Based on the target association cost value corresponding to the sensor set at the multiple times, the trend of association cost change is determined;

[0017] If the trend of the associated cost change is greater than or equal to a first threshold, each sensor in the sensor set is identified as a target sensor.

[0018] Optionally, the target association value includes the first association value between the first sensor and a second sensor other than the first sensor in the sensor set. Determining the target sensor from the sensor set based on the target association values ​​corresponding to the sensor set at the plurality of times includes:

[0019] Determine the first quantity corresponding to the first associated cost value of the first sensor that is less than or equal to the second threshold;

[0020] If the first quantity is greater than or equal to the second quantity, the first sensor is determined as the target sensor; wherein, the second quantity is obtained by subtracting the number of sensors in the sensor set from a preset quantity;

[0021] When the first number is less than or equal to the third number, the sensor set is reconstructed based on the first sensor, and the first target detection results corresponding to each sensor in the reconstructed sensor set are fused to update the first fused target; wherein the third number is less than the second number.

[0022] Optionally, determining the first association value between the first sensor and a second sensor other than the first sensor in the sensor set based on the first target detection result includes:

[0023] Based on the first target detection result of the first sensor and the first target detection result of the second sensor, the target Mahalanobis distance and target state similarity are determined;

[0024] The first association value is obtained by weighted summation of the target Mahalanobis distance and the target state similarity.

[0025] Optionally, determining the first association value between the first sensor and a second sensor other than the first sensor in the sensor set based on the first target detection result includes:

[0026] The velocity parameters of the first target are determined based on the first target detection result of the first sensor.

[0027] The velocity parameters of the second target are determined based on the first target detection result of the second sensor;

[0028] The first associated cost value is determined based on the difference between the first target velocity parameter and the second target velocity parameter.

[0029] Optionally, the method further includes:

[0030] If the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than a preset distance, then a second fusion target within a preset range of the first fusion target is obtained;

[0031] Based on the first target detection result, determine the second association value between the third sensor of the first sensor type in the sensor set and the fourth sensor of the second sensor type in the sensor set;

[0032] Based on the first target detection result and the second target detection result corresponding to the second fused target, a third association generation value between the third sensor and the fourth sensor is determined;

[0033] If the third association cost is less than the second association cost, a fusion association relationship is established between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor, so as to fuse the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor to obtain the third fused target.

[0034] Secondly, this application provides a target detection device, the device comprising:

[0035] The acquisition module is used to acquire a set of first target detection results corresponding to the first fusion target at multiple time points; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set.

[0036] The cost value module is used to determine the target association cost value of the sensor set consisting of the sensors corresponding to the first target detection results, based on the first target detection result set.

[0037] The sensor module is used to determine the target sensor from the sensor set based on the target association value corresponding to the sensor set at the multiple times.

[0038] The detection module is used to independently determine the first target based on the first target detection result of the target sensor.

[0039] Optionally, the cost module includes:

[0040] The first association value submodule is used to determine the first association value between the first sensor and the second sensor other than the first sensor in the sensor set based on the first target detection result.

[0041] The target association value submodule is used to determine the target association value corresponding to the sensor set based on the first association value and the number of sensors in the sensor set.

[0042] Optionally, the sensor module includes:

[0043] The association cost change trend submodule is used to determine the association cost change trend based on the target association cost value corresponding to the sensor set at the multiple times.

[0044] The target sensor submodule is used to identify each sensor in the sensor set as a target sensor when the trend of the associated cost change is greater than or equal to a first threshold.

[0045] Optionally, the target association value includes the first association value between the first sensor and a second sensor other than the first sensor in the sensor set, and the sensor module includes:

[0046] The first quantity submodule is used to determine the first quantity corresponding to the first associated cost value of the first sensor that is less than or equal to the second threshold.

[0047] A target sensor determination submodule is used to determine the first sensor as the target sensor when the first quantity is greater than or equal to the second quantity; wherein the second quantity is obtained by subtracting the number of sensors in the sensor set from a preset quantity;

[0048] The fusion target update submodule is used to reconstruct the sensor set based on the first sensor when the first number is less than or equal to the third number, and to fuse the first target detection results corresponding to each sensor in the reconstructed sensor set to update the first fusion target; wherein the third number is less than the second number.

[0049] Optionally, the first associated cost submodule includes:

[0050] The similarity unit is used to determine the target Mahalanobis distance and target state similarity based on the first target detection result of the first sensor and the first target detection result of the second sensor;

[0051] The first association value calculation unit is used to perform a weighted summation of the target Mahalanobis distance and the target state similarity to obtain the first association value.

[0052] Optionally, the first associated cost submodule includes:

[0053] The first target velocity parameter unit is used to determine the first target velocity parameter based on the first target detection result of the first sensor;

[0054] The second target velocity parameter unit is used to determine the second target velocity parameter based on the first target detection result of the second sensor;

[0055] The first associated cost calculation unit is used to determine the first associated cost based on the difference between the first target speed parameter and the second target speed parameter.

[0056] Optionally, the device further includes:

[0057] The second fusion target module is used to acquire a second fusion target within a preset range of the first fusion target when the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than a preset distance.

[0058] The second association value module is used to determine the second association value between the third sensor of the first sensor type in the sensor set and the fourth sensor of the second sensor type in the sensor set based on the first target detection result.

[0059] The third association value module is used to determine the third association value between the third sensor and the fourth sensor based on the first target detection result and the second target detection result corresponding to the second fused target.

[0060] The third fusion target module is used to establish a fusion association relationship between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor when the third association cost is less than the second association cost, so as to fuse the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor to obtain the third fusion target.

[0061] Thirdly, this application provides a vehicle controller, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0062] Fourthly, this application provides a readable storage medium that, when the instructions in the readable storage medium are executed by the processor of a vehicle controller, enables the vehicle controller to perform the steps of the method described in the first aspect.

[0063] Fifthly, embodiments of this application provide an electronic device including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions being executed by the processor to implement the steps of the method described in the first aspect.

[0064] Sixthly, this application provides a vehicle including the aforementioned vehicle controller.

[0065] Compared with existing technologies, the target detection method, device, vehicle controller, and vehicle described in this application have the following advantages:

[0066] In summary, this application provides a target detection method, comprising: acquiring a set of first target detection results corresponding to a first fusion target at multiple time points; wherein the first fusion target is obtained by fusing each first target detection result in the set of first target detection results; determining the target association value of a sensor set consisting of sensors corresponding to the first target detection results based on the set of first target detection results; determining a target sensor from the sensor set based on the target association value corresponding to the sensor set at multiple time points; and independently determining a first target through the first target detection results of the target sensor. This method can detect the target fusion results corresponding to the sensors that generate the fusion target, promptly identify sensors unsuitable for further generating the fusion target, and thus independently perform target detection on the target detection results of these sensors, avoiding the impact of these sensor target detection results on the accuracy of the fusion target, thereby improving the accuracy of target detection. Attached Figure Description

[0067] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0068] Figure 1 A flowchart of a target detection method provided in this application embodiment;

[0069] Figure 2The present application provides a schematic diagram of target detection result fusion.

[0070] Figure 3 A flowchart illustrating another target detection method provided in this application embodiment;

[0071] Figure 4 A schematic diagram of a preset range provided for an embodiment of this application;

[0072] Figure 5 This is a structural block diagram of a target detection device provided in an embodiment of this application. Detailed Implementation

[0073] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0074] The present application will now be described in detail with reference to the accompanying drawings and embodiments.

[0075] Reference Figure 1 The diagram shows a flowchart of a target detection method provided in an embodiment of this application.

[0076] Step 101: Obtain the first target detection result set corresponding to the first fusion target at multiple times; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set.

[0077] Object detection is a core task in computer vision, referring to the detection of the location, category, and other information of target objects in images or videos. Object detection algorithms are mostly based on deep learning, especially convolutional neural networks (CNNs). Common object detection algorithms include, but are not limited to, YOLO (You Only Look Once), Faster R-CNN, and SSD (Single Shot MultiBox Detector). Object detection has wide applications, such as in autonomous driving technology to identify and locate pedestrians, vehicles, traffic signs, traffic lights, and obstacles on roads; in video surveillance and security to identify abnormal behavior and intrusion detection; and in drone navigation to identify and locate objects around the drone, perform obstacle recognition, and object grasping.

[0078] Target detection can be performed based on a single sensor or multiple sensors to obtain a fused target. The fused target, in a multi-sensor target detection system, combines perception data from multiple sensors on the same object or environment to form a more accurate and complete target information. These sensors may include, but are not limited to, cameras, millimeter-wave radars, LiDAR (Light Detection and Ranging), ultrasonic sensors, and infrared sensors; the embodiments in this application do not impose specific limitations.

[0079] In target detection fusion, the target detection results from different sensors can be compared and analyzed to match these results and determine whether they originate from the same real target. This allows for the fusion of target detection results from multiple sensors corresponding to the same real target, resulting in a fused target. The matching methods described above may include, but are not limited to, ID matching, location matching, feature matching, etc., and are not specifically limited in the embodiments of this application.

[0080] In this embodiment, the first fusion target represents the target detection result obtained by fusing matching first target detection results determined by multiple sensors. To ensure the accuracy of the first fusion target generated by fusing multiple sensors, the first fusion target can be verified. First, the set of first target detection results corresponding to the first fusion target at multiple times can be obtained; wherein, the first fusion target is obtained by fusing each first target detection result in the set of first target detection results.

[0081] For example, the aforementioned multiple times can be the times corresponding to the first 3 target detections. Then, the first fused target at time T0 is obtained by fusing the first target detection result A0 of sensor A and the first target detection result B0 of sensor B. The first fused target at time T-1 is obtained by fusing the first target detection result A1 of sensor A and the first target detection result B1 of sensor B. The first fused target at time T-2 is obtained by fusing the first target detection result A2 of sensor A and the first target detection result B2 of sensor B. The sets of first target detection results corresponding to the first fused target at the multiple times are as follows: (first target detection result A0, first target detection result B0), (first target detection result A1, first target detection result B1), (first target detection result A2, first target detection result B2).

[0082] It should be noted that the above-mentioned multiple moments can be determined by preset time rules. For example, the preset time rules can be all update moments of the first fusion target in the past 1 second, the update moments of the first fusion target in the past 10 times, etc. This application embodiment does not make specific limitations.

[0083] Step 102: Based on the first target detection result set, determine the target association value of the sensor set consisting of the sensors corresponding to the first target detection results.

[0084] In this embodiment, for each time step corresponding to the set of first target detection results, the target association cost value of the sensor set consisting of the sensors corresponding to each first target detection result can be determined based on these results. The association cost value is a numerical indicator used in multi-sensor fusion to measure whether targets observed by different sensors belong to the same object. It is evaluated by calculating the differences between sensor data (first target detection results). The smaller the association cost value, the closer the first target detection results observed by the two sensors are, and the more likely they are to be the same entity; the larger the cost value, the greater the difference between the data, the lower the association, and the more likely they are to be different entities.

[0085] Specifically, based on the set of first target detection results at a given time, the target association value corresponding to the sensor set, which comprises all the sensors corresponding to the first target detection results included in the set at that time, can be calculated. In the embodiments of this application, the target association value corresponding to the sensor set can represent the first association value between any two sensors in the sensor set, or it can represent the association value calculated based on the first association value between any two sensors in the sensor set. This embodiment of the application does not impose any specific limitation.

[0086] In one implementation, the target association cost can be obtained by summing or averaging the first association cost values ​​between the first target detection results corresponding to any two sensors in the sensor set. In another implementation, the average association cost value between the first target detection results corresponding to any two sensors in the sensor set can be calculated first, and then the target association cost value can be determined based on the degree of difference between the first target detection results corresponding to each sensor in the sensor set and the average association cost value. In yet another implementation, the first association cost value between sensors of a first sensor type and sensors of other sensor types in the sensor set can be calculated, and the target association cost value can be obtained by summing or averaging these first association cost values, thereby reducing the number of first association cost values ​​that need to be determined and improving computational efficiency. The first sensor type can be a type of sensor with high accuracy, such as a lidar sensor.

[0087] In this application embodiment, the calculation of the first association cost value can be based on multiple different dimensions, such as the target's position, speed, shape features, time, etc. Different calculation methods are suitable for different application scenarios. For example, for moving targets (such as pedestrians, vehicles, etc.), in addition to position, speed can also be an important matching criterion. The speed difference between speed measurements of the same moving target by different sensors can be compared. The smaller the speed difference, the smaller the cost value, indicating that the target's motion state is similar and the probability of them being the same target is higher, and vice versa. For targets with specific appearances, such as object images, contours, colors, etc. provided by cameras or LiDAR, the association cost value can be calculated by comparing the target contour information, target shape information, and other features in the first target detection results corresponding to the sensors. The smaller the feature difference, the smaller the cost value, indicating that the target features observed by the sensors are similar and the probability of them being the same target is higher, and vice versa. This application embodiment does not specifically limit the calculation method of the association cost value, and technicians can flexibly adjust it according to actual business needs.

[0088] Step 103: Based on the target association value corresponding to the sensor set at the multiple times, determine the target sensor from the sensor set.

[0089] In this embodiment, the target sensor can be determined from the sensor set based on the target association cost value corresponding to the first fusion target at multiple times. Here, the target sensor represents the sensor whose association needs to be de-associated, meaning that the first target detection result corresponding to the target sensor is no longer fused with other sensors in the sensor set to independently determine the first target.

[0090] Specifically, when the target association cost value is the association value calculated based on the first association cost value between each pair of sensors in the sensor set, the number of first-generation values ​​of the target association cost value that is greater than or equal to the first association cost threshold can be determined. If the number of first-generation values ​​exceeds the first preset number, each target sensor in the sensor set is directly determined as a target sensor. Alternatively, based on the number of first-generation values ​​of the target association cost value that is greater than or equal to the first association cost threshold and the total number of target association cost values, the proportion of target association cost values ​​that are greater than or equal to the first association cost threshold can be calculated. If the proportion is greater than or equal to the preset proportion, each target sensor in the sensor set is directly determined as a target sensor.

[0091] When the target association cost includes the first association cost based on the sensor pairs formed by two sensors in the sensor set, the number of second-generation values ​​of the first association cost that are greater than or equal to the second association cost threshold can be determined among the first association cost corresponding to each sensor pair at multiple times. If the number of second-generation values ​​is greater than or equal to the second preset number, the association relationship between the sensors in the sensor pair is canceled; otherwise, the association relationship between the sensors in the sensor pair is retained. After performing the above operation on each sensor pair in the sensor set, the target sensor in the sensor set that is not associated with other sensors is obtained.

[0092] For example, if there exists a sensor set [sensor 1, sensor 2, sensor 3, sensor 4], the conclusions regarding whether to retain or cancel the association relationship between the sensor pairs formed by each sensor are shown in Table 1 below:

[0093] Sensor 1 Sensor 2 Sensor 3 Sensor 4 Sensor 1 reserve reserve Cancel Sensor 2 Cancel Cancel Sensor 3 Cancel Sensor 4

[0094] Table 1

[0095] In this process, sensor 4 is de-associated with all other sensors, sensor 2 retains its association with sensor 1, sensor 2 is de-associated with other sensors, and sensor 3 retains its association with sensor 1. Therefore, sensor 4 can be identified as the target sensor, allowing the first target detection result acquired by sensor 4 to independently determine the target without participating in target fusion. Furthermore, since there are direct or indirect associations between sensors 1, 2, and 3, the first target detection results from sensors 1, 2, and 3 can be further fused to determine the first fusion target. This eliminates sensor 4, which has a larger error margin, from the fusion information source of the first fusion target, improving the accuracy of the first fusion target.

[0096] Step 104: Independently determine the first target based on the first target detection result of the target sensor.

[0097] After identifying the target sensor from the sensor set, the fusion operation of each target sensor for the first target detection result can be deactivated in the subsequent target detection process, and the first target can be independently determined using the first target detection result of the target sensor.

[0098] Furthermore, if some sensors in the sensor set are identified as target sensors, these target sensors can independently determine the target, and the first target detection results of the remaining sensors can still participate in target fusion to obtain a fused target. For example, if there is a sensor set [sensor 1, sensor 2, sensor 3, sensor 4], where sensor 1 and sensor 2 are identified as target sensors, then the first target 1 is independently determined based on the first target detection result of sensor 1, the first target 2 is independently determined based on the first target detection result of sensor 2, and the first target detection results of sensor 3 and sensor 4 are fused to obtain the fused target 1.

[0099] Reference Figure 2 , Figure 2 This illustration shows a schematic diagram of target detection result fusion provided in an embodiment of this application, such as... Figure 2 As shown, the first fusion target is obtained by fusing the first target detection result 21 identified in the data D1 collected by sensor 1, the first target detection result 22 identified in the data D2 collected by sensor 2, and the first target detection result 23 identified in the data D3 collected by sensor 3. At time T0, the target association cost value of the first target detection result 1-3 of the sensor set consisting of sensors 1-3 is N1; at time T1, the target association cost value of the first target detection result 1-3 of the sensor set consisting of sensors 1-3 is N2; at time T2, the target association cost value of the first target detection result 1-3 of the sensor set consisting of sensors 1-3 is N3. If N1, N2, and N3 are all greater than the first association cost threshold, then each of sensors 1 to 3 can be directly identified as a target sensor, and at time T3, the first target detection result of each sensor can be independently identified as the corresponding first target.

[0100] In summary, this application provides a target detection method, comprising: acquiring a set of first target detection results corresponding to a first fusion target at multiple time points; wherein the first fusion target is obtained by fusing each first target detection result in the set of first target detection results; determining the target association value of a sensor set consisting of sensors corresponding to the first target detection results based on the set of first target detection results; determining a target sensor from the sensor set based on the target association value corresponding to the sensor set at multiple time points; and independently determining a first target through the first target detection results of the target sensor. This method can detect the target fusion results corresponding to the sensors that generate the fusion target, promptly identify sensors unsuitable for further generating the fusion target, and thus independently perform target detection on the target detection results of these sensors, avoiding the impact of these sensor target detection results on the accuracy of the fusion target, thereby improving the accuracy of target detection.

[0101] Reference Figure 3 , Figure 3A flowchart of another target detection method provided in an embodiment of this application is shown.

[0102] Step 201: Obtain the set of first target detection results corresponding to the first fusion target at multiple times; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set.

[0103] This step can be found in step 101, and will not be repeated in this embodiment.

[0104] Step 202: Based on the first target detection result, determine the first association value between the first sensor and the second sensor other than the first sensor in the sensor set.

[0105] In this embodiment, each sensor in the sensor set can serve as a first sensor, and the second sensor corresponding to the first sensor is any other sensor in the sensor set besides the first sensor. The first sensor can form a sensor pair with each corresponding second sensor, and a first association value is calculated between the first sensor and any second sensor other than the first sensor.

[0106] Optionally, the first associated cost value can be determined based on the following sub-steps 2021 to 2022, where step 202 may include:

[0107] Sub-step 2021: Based on the first target detection result of the first sensor and the first target detection result of the second sensor, determine the target Mahalanobis distance and the target state similarity.

[0108] In this embodiment of the application, the target Mahalanobis distance and target state similarity between the first target detection results of the first sensor and the corresponding first target detection results of the second sensor can be calculated.

[0109] The target Mahalanobis distance represents the Mahalanobis distance between the two targets indicated by the first target detection results, and can be calculated using the following formula 1:

[0110] DM = (x1 - x2) T S -1 (x1-x2) Formula 1

[0111] Where DM represents the Mahalanobis distance between the targets, x1 and x2 represent the feature vectors of the two first target detection results, and S is the covariance matrix of the features, representing the correlation between the features. -1 The inverse of the covariance matrix reflects the influence between features, and T represents the transpose operation.

[0112] Target state similarity represents the similarity between target states such as target orientation, target category, target size, target height, target width, and target position in two first target detection results. Target state similarity can represent the similarity between a single target state or can be calculated based on multiple similarities between various target states; this application does not impose specific limitations on this. The similarity between target states can be calculated using normalized state feature vectors, for example, by using cosine similarity or Euclidean distance to quantify the similarity between these features.

[0113] For example, target location information and target orientation information can be extracted from two first target detection results, the positional similarity between the two target locations and the orientational similarity between the two target orientations can be calculated, and then the positional similarity and orientational similarity can be averaged or weighted to obtain the target state similarity between the two first target detection results.

[0114] Sub-step 2022: The target Mahalanobis distance and the target state similarity are weighted and summed to obtain the first association value.

[0115] In this embodiment of the application, the target Mahalanobis distance and the target state similarity can be weighted and summed to obtain the first association value.

[0116] Furthermore, due to different target categories, the impact of Mahalanobis distance and state similarity on the association value varies. For example, for targets such as pedestrians and non-motorized vehicles, the accuracy of the detected target state is often low. In this case, Mahalanobis distance should be given more weight, with a higher weight assigned to the target Mahalanobis distance and a lower weight assigned to the target state similarity. On the other hand, for targets such as cars and trucks that are larger and have more regular structures, the detected target state is often more accurate. In this case, state similarity should be given more weight, with a lower weight assigned to the target Mahalanobis distance and a higher weight assigned to the target state similarity.

[0117] Therefore, in this embodiment, when the target types contained in the two first target detection results are consistent, the weights corresponding to the target Mahalanobis distance and target state similarity can be determined according to the target type, thereby making the first association value obtained by weighted summation more accurate. The above weight determination method may include table lookup, cloud query, etc.

[0118] By combining the target detection results from the first and second sensors, the Mahalanobis distance and target state similarity are determined. A weighted sum of the Mahalanobis distance and target state similarity yields the first association cost. The Mahalanobis distance considers the overall correlation between target detection results, while target state similarity focuses on the similarity of specific features (such as direction and shape). Combining these two indicators reduces the risk of misjudgment caused by a single indicator, provides a more comprehensive assessment, and more meticulously reflects the relationship between the first target detection results, thus improving the accuracy of determining the association cost.

[0119] Optionally, the first associated cost value can be determined based on the following sub-steps 2023 to 2025, where step 202 may include:

[0120] Sub-step 2023: Determine the first target velocity parameter based on the first target detection result of the first sensor.

[0121] In this embodiment, a first target velocity parameter can be determined based on the first target detection result of a first sensor. The first target velocity parameter may include at least one of the following: target velocity, change in target velocity, target acceleration, change in target acceleration, etc., and this embodiment does not impose specific limitations on these parameters.

[0122] It should be noted that since a single first target detection result can only reflect the information of the target at a certain point in time, while the first target velocity parameter needs to be calculated based on the orientation information, position information, etc. at multiple points in time, the first target velocity parameter can be determined based on the first target detection results of the first sensor at two or more times.

[0123] Sub-step 2024: Determine the velocity parameters of the second target based on the first target detection result of the second sensor.

[0124] Sub-step 2025: Determine the first associated cost value based on the difference between the first target velocity parameter and the second target velocity parameter.

[0125] In this embodiment, a first correlation coefficient value between the detection results of the two sensors for the two first targets can be determined based on the difference between the first target velocity parameter corresponding to the first sensor and the second target velocity parameter corresponding to the second sensor. The difference may include, but is not limited to, absolute value differences, proportional differences, etc., and this application does not impose specific limitations on it.

[0126] Specifically, the first associated cost value can be obtained by calculating the proportional differences between parameters of the same type in the first and second target speed parameters, and then averaging or weighted averaging all proportional differences. Alternatively, the first associated cost value can be obtained by calculating the absolute differences between parameters of the same type in the first and second target speed parameters, and then averaging or weighted averaging all absolute differences. Furthermore, the first associated cost value can be obtained by calculating the proportional differences between some parameters of the same type in the first and second target speed parameters, and the absolute differences between other parameters of the same type, and then weighted averaging or weighted averaging the aforementioned proportional and absolute differences.

[0127] For example, if the first target velocity parameter of the first sensor is [velocity 1, acceleration 1], and the second target velocity parameter of the second sensor is [velocity 2, acceleration 2], then the absolute difference 'a' between velocity 1 and velocity 2, and the ratio 'b' between acceleration 1 and acceleration 2 can be calculated. Then, the first associated cost 'c' can be calculated based on the following formula 2:

[0128] c = k1*a + k2*b (Formula 2)

[0129] Wherein, k1 represents the velocity weight and k2 represents the acceleration weight. k1 and k2 can be flexibly set according to actual business needs, and this application embodiment does not impose specific limitations.

[0130] A first target velocity parameter is determined based on the first target detection result from a first sensor; a second target velocity parameter is determined based on the first target detection result from a second sensor; and a first correlation cost value is determined based on the difference between the first and second target velocity parameters. This method allows for the determination of target velocity parameters from continuous first target detection results over a time series, and then the determination of the correlation cost value based on these parameters. This enables the correlation cost value to assess the difference between the target detection results from the two sensors over a period of time, thus improving the accuracy of determining the correlation cost value.

[0131] Step 203: Based on the first association value and the number of sensors in the sensor set, determine the target association value corresponding to the sensor set.

[0132] In this embodiment, the target association cost of the sensor set can be determined based on the first association cost of each sensor pair in the sensor set and the number of sensors in the sensor set. Specifically, the first association costs of each sensor pair can be summed, and the summed result can be compared with the number of sensors to obtain the target association cost of the sensor set. Alternatively, the number of sensor pairs that can be formed in the sensor set can be calculated based on the number of sensors in the sensor set, and the summed result can be compared with the number of sensor pairs to obtain the target association cost of the sensor set.

[0133] Furthermore, increasing the number of sensors in the sensor set increases the difficulty of sensor correlation. Even with a higher correlation statistic, the accuracy of fused target detection can still be improved to some extent. Therefore, when the number of sensors is high, the requirement for the correlation statistic value can be appropriately relaxed. In this embodiment, the average correlation statistic value of the first correlation statistic value of each sensor pair in the sensor set can be calculated. Then, the average correlation statistic value is corrected according to the number of sensors in the sensor set to obtain the target correlation statistic value. This ensures that, with the same average correlation statistic value, the higher the number of sensors, the lower the corrected target correlation statistic value, and vice versa.

[0134] Based on the first target detection result, the first association cost value between the first sensor and a second sensor (excluding the first sensor) in the sensor set is determined. Based on the first association cost value and the number of sensors in the sensor set, the target association cost value corresponding to the sensor set is determined. This method considers the first association cost value of all sensor pairs and the number of sensors when calculating the target association cost value of the sensor set, resulting in a more refined calculation process and improving the accuracy of determining the target association cost value of the sensor set.

[0135] Step 204: Determine the trend of association cost change based on the target association cost corresponding to the sensor set at the multiple times.

[0136] In this embodiment, the trend of association cost change can also be determined based on the target association cost value corresponding to the sensor set at multiple times. The trend of association cost change can be determined based on one or more indicators that can describe the trend of change, such as the amount of change, the magnitude of change, and the standard deviation, and this embodiment does not impose specific limitations. The amount of change can include a decrease or an increase, and the magnitude of change can include an increase or a decrease.

[0137] Specifically, the target association cost value corresponding to the sensor set at multiple times can be calculated, and the change in each type of indicator in the target association cost value can be calculated. These changes can be weighted and averaged to obtain the trend of association cost change.

[0138] Furthermore, by inputting the target association costs corresponding to multiple time points into a time series model, the changing trend of the association costs output by the time series model can be obtained. The time series model can extract the changing characteristics of the time series, thereby outputting the changing trend of the data. The aforementioned time series model may include, but is not limited to, Long Short-Term Memory Networks (LSTM), Autoregressive Moving Average Models (ARMA), etc., and the embodiments of this application do not impose specific limitations.

[0139] Step 205: If the trend of the associated cost change is greater than or equal to the first threshold, each sensor in the sensor set is identified as a target sensor.

[0140] In this embodiment, a first threshold can be preset. If the trend of the associated cost change is greater than or equal to the first threshold, it is determined that the sensors included in the sensor set meet the splitting conditions. The first target detection results corresponding to each sensor can be split and target detection can be performed independently. The first threshold can be flexibly set by technicians according to actual business needs, and this embodiment does not impose specific limitations.

[0141] By determining the trend of association cost changes based on the target association cost value corresponding to the sensor set at multiple times, and when the trend of association cost changes is greater than or equal to a first threshold, each sensor in the sensor set is identified as a target sensor. This allows for timely and accurate determination of whether the sensor set needs to cancel the association for the first target detection result, and enables independent target detection for the first target detection results of each sensor, which helps to further reduce the error rate of fused target detection.

[0142] Optionally, the target association cost may include the first association cost between the first sensor and a second sensor other than the first sensor in the sensor set. That is, the target association cost of the sensor set can be directly composed of the first association cost between the first sensor and the second sensor in the sensor set. Then, the target sensor can be determined through the following steps A1 to A2:

[0143] Step A1: Determine the first quantity corresponding to the first associated cost value of the first sensor that is less than or equal to the second threshold.

[0144] In this embodiment, a first number of first association values ​​of the first sensors that are less than or equal to a second threshold can be determined firstly. That is, for each first sensor, first association values ​​less than or equal to the second threshold are selected from all its corresponding first association values, and then the first number of these selected first association values ​​is counted. This first number represents the number of second sensors that have a lower degree of matching with the first target detection result of the first sensor.

[0145] Step A2: If the first quantity is greater than or equal to the second quantity, the first sensor is determined as the target sensor; wherein the second quantity is obtained by subtracting the number of sensors in the sensor set from a preset quantity.

[0146] In this embodiment, a second quantity can be calculated based on the number of sensors in the sensor set and a preset quantity. Specifically, the second quantity can be obtained by subtracting the preset quantity from the number of sensors.

[0147] If the first quantity corresponding to a certain first sensor is greater than or equal to the second quantity, it means that the first target detection result of the first sensor does not match the first target detection results of many sensors. The first target detection result of the first sensor is not suitable to continue to participate in target fusion. Therefore, the connection between the first sensor and the corresponding second sensor can be cancelled, and the first sensor can be identified as a target sensor. Thus, independent target detection can be performed based on the first target detection result of the first sensor.

[0148] A3, when the first number is less than or equal to the third number, the sensor set is reconstructed based on the first sensor, and the first target detection results corresponding to each sensor in the reconstructed sensor set are fused to update the first fused target; wherein, the third number is less than the second number.

[0149] In this embodiment, a third quantity can be set, such that the third quantity is less than the second quantity. If the first quantity corresponding to a certain first sensor is less than or equal to the third quantity, it indicates that the first target detection result of the first sensor matches the first target detection results of a large number of sensors, and the first target detection result of the first sensor is suitable to continue participating in target fusion. In this case, the first sensor can be retained in the sensor set, and a sensor set for fusing to obtain the first fused target can be reconstructed based on all first sensors that meet this condition. The first target detection results of the sensors in the reconstructed sensor set participate in target fusion, the first fused target is updated, and a more accurate first fused target is obtained.

[0150] For example, if the detection result of the first target participating in the first fusion target comes from sensor set A [sensor a, sensor b, sensor c, sensor d], then sensor a as the second target sensor corresponding to the first sensor includes sensor b, sensor c, and sensor d; sensor b as the second target sensor corresponding to the first sensor includes sensor a, sensor c, and sensor d, and so on. The number of each sensor in the above sensor set A corresponding to the first sensor can be shown in Table 2 below:

[0151] First sensor First quantity Sensor a 1 Sensor b 1 sensor c 1 sensor d 3

[0152] Table 2

[0153] Given that the second quantity is 2 and the third quantity is 1, since the first quantity of sensors a, b, and c is equal to the third quantity, sensors a, b, and c are retained in the sensor set A. Since the first quantity of sensor d is greater than the second quantity, sensor d is removed from the sensor set A and used as the target sensor to independently determine the first target based on its first target detection result. Using sensors a, b, and c, a new sensor set B [sensor a, sensor b, sensor d] is constructed. The first target detection results corresponding to the three sensors in the reconstructed sensor B can be used for target fusion to obtain a more accurate first fused target.

[0154] By determining a first number corresponding to a first association cost value of a first sensor that is less than or equal to a second threshold, and if the first number is greater than or equal to a second number, the first sensor is identified as the target sensor. This method allows for the rapid identification of sensors unsuitable for further association by comparing the magnitude of the first association cost values, quickly separating sensors from the sensor set that are no longer suitable for target fusion, thus improving the efficiency of target sensor identification. Furthermore, by excluding target sensors unsuitable for target fusion from the sensor set, interference from target sensors on the first fusion target is eliminated, which helps improve the accuracy of the first fusion target.

[0155] Optionally, in embodiments of this application, the following steps B1 to B4 may also be included:

[0156] Step B1: If the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than the preset distance, obtain the second fusion target within the preset range of the first fusion target.

[0157] Because different sensors have different detection success rates for certain types of targets at different distances, some targets may be missed at long distances. As the distance increases, the sensor's target detection accuracy also increases, enabling it to identify targets missed at long distances. This can lead to the establishment of correct correlations for target detection results for distant targets. However, as the target gets closer, the sensor may detect higher-quality correlated targets in the vicinity of the target, which may cause the target to adopt an incorrect first target detection result, resulting in a decrease in the accuracy of the target fusion.

[0158] Therefore, in this embodiment, the target type and target distance of the first fused target can be determined. If the first fused target belongs to a preset target type and the target distance of the first fused target is less than the preset distance, the target detection result of the fused target can be re-determined. First, a second fused target within a preset range of the first fused target can be obtained. Here, the preset range represents the range of the first fused target in three-dimensional space (e.g., world coordinate system, vehicle coordinate system). This preset range can be a spherical range or a rectangular range, and this embodiment does not specifically limit it.

[0159] Reference Figure 4 , Figure 4 This application provides a schematic diagram of a preset range, as shown in the embodiment. Figure 4 As shown, the target preset range 42 of the first fusion target 41 in the world coordinate system 40 may include the second fusion target 43 and the second fusion target 44.

[0160] The aforementioned preset target types may include, but are not limited to, one or more of the following: fusion type, scene type, and object type. Among them, object type may include, but is not limited to, pedestrian type, bicycle type, etc.; fusion type may include, but is not limited to, camera fusion type (i.e., fusion using only the first target detection result corresponding to the camera), radar and camera fusion type (i.e., fusion using the first target detection result corresponding to the camera and the first target detection result corresponding to the radar); scene type may include, but is not limited to, low-speed static scene type (e.g., traffic light waiting scene), high-speed dense scene type (e.g., busy expressway scene), low-speed dense scene type (e.g., congestion scene), etc.

[0161] It should be noted that the preset target type can be a collection of the aforementioned types. If the first fused target matches each type in the preset target type, it can be determined that the first fused target belongs to the preset target type. For example, the preset target type may include pedestrian type, camera fusion type, and low-speed static scene type. If a first fused target is a pedestrian in a low-speed static scene and is obtained by fusing the first target detection results of two cameras, then the first fused target belongs to the aforementioned preset target type.

[0162] In this embodiment of the application, when the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than the preset distance, the target detection results of the sensors of the first fusion target and the surrounding second fusion target can be recombined to re-determine the target detection result combination that can achieve a better fusion effect. When the first fusion target does not meet the above conditions, only the target detection results corresponding to the first fusion target can be processed, and the target sensor can be determined from the sensor set corresponding to the first fusion target using the method in the above steps.

[0163] Step B2: Based on the first target detection result, determine the second association value between the third sensor of the first sensor type in the sensor set and the fourth sensor of the second sensor type in the sensor set.

[0164] In this embodiment, a second association value can be determined between a third sensor of a first sensor type in the sensor set and a fourth sensor of a second sensor type in the sensor set based on the first target detection result. The first sensor type can represent a sensor type with high target detection accuracy, such as lidar, while the second sensor type can represent a sensor type with lower target detection accuracy compared to the first sensor type, such as ultrasonic radar.

[0165] Specifically, a third sensor of a first sensor type and a fourth sensor of a second sensor type can be selected from the sensor set. Based on the first target detection results corresponding to each third sensor and each fourth sensor used to fuse the first fusion target, a second association cost value is determined between each pair of third and fourth sensors. The calculation method for the second association cost value can refer to the calculation method for the first association cost value described above, and will not be repeated in this embodiment.

[0166] Step B3: Based on the first target detection result and the second target detection result corresponding to the second fused target, determine the third association generation value between the third sensor and the fourth sensor.

[0167] In this embodiment of the application, the third association generation value between the third sensor and the fourth sensor can be determined based on the first target detection result corresponding to the first fusion target and the second target detection result corresponding to the second fusion target.

[0168] Specifically, the third correlation generation value between the third sensor and the fourth sensor can be calculated based on the first target detection result of the third sensor participating in the fusion to obtain the first fusion target, and the second target detection result of the fourth sensor participating in the fusion to obtain the second fusion target.

[0169] For example, if there are a first fusion target T1, a second fusion target T2, and a second fusion target T3, wherein the first fusion target T1 is obtained by fusing the first target detection results O1 of the third sensor S1, O2 of the third sensor S2, and O3 of the third sensor S3; the second fusion target T2 is obtained by fusing the second target detection results O4 of the fourth sensor S2, O5 of the fourth sensor S3, and O6 of the fourth sensor S4; and the second fusion target T3 is obtained by fusing the second target detection results O7 of the fourth sensor S1 and O8 of the fourth sensor S4, then for each pair of first and second fusion targets, based on the first target detection results of each third sensor corresponding to the first fusion target and the second target detection results of each fourth sensor corresponding to the second fusion target, multiple sets of third association generation values ​​between the third and fourth sensors can be calculated. The target detection results used for the 12 third association generation values ​​calculated based on the above data can be shown in Table 3 below:

[0170]

[0171]

[0172] Table 3

[0173] Step B4: If the third association cost is less than the second association cost, establish a fusion association relationship between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor, so as to fuse the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor to obtain the third fused target.

[0174] In this embodiment, when the third association value is less than the second association value, a fusion association relationship can be established between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor, thereby achieving target fusion of the first target detection result and the second target detection result to obtain the third fused target.

[0175] For example, in Table 3 above, if the second association cost value between the third sensor S1 and the third sensor S2 calculated from the first target detection result O1 of the third sensor S1 and the first target detection result O2 of the third sensor S2 is 0.6, the third association cost value 1 between the first target detection result O1 of the third sensor S1 and the second target detection result O4 of the third sensor S2 is 0.7, and the third association cost value 1 between the first target detection result O1 of the third sensor S1 and the second target detection result O4 of the third sensor S2 is 0.5, then the fusion association relationship between the first target detection result O1 of the third sensor S1 and the first target detection result O2 of the third sensor S2 can be cancelled, and the first target detection result O1 of the third sensor S1 can be re-established. The fusion correlation between the first target detection result O1 and the second target detection result O4 of the third sensor S2 is established. Furthermore, if the third correlation value 2 between the first target detection result O1 and the second target detection result O5 is less than the second correlation value between the first target detection result O1 and the first target detection result O3, and the third correlation value 5 between the first target detection result O2 and the second target detection result O5 is less than the second correlation value between the first target detection result O2 and the first target detection result O3, then a fusion correlation relationship between the first target detection result O1, the second target detection result O4 and the second target detection result O5 can be established, thereby fusing the first target detection result O1, the second target detection result O4 and the second target detection result O5 to obtain the third fused target.

[0176] In this embodiment, when the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than the preset distance, the fusion correlation between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor can be re-established based on the correlation value between the target detection results corresponding to the first fusion target and the surrounding second fusion targets, which helps to improve the fusion accuracy of certain close-range fusion targets.

[0177] In summary, this application provides another target detection method, including: obtaining a set of first target detection results corresponding to a first fusion target at multiple time points; wherein the first fusion target is obtained by fusing each first target detection result in the first target detection result set; determining the target association value of a sensor set composed of sensors corresponding to the first target detection results based on the first target detection result set; determining a target sensor from the sensor set based on the target association value corresponding to the sensor set at multiple time points; and independently determining the first target through the first target detection result of the target sensor. This method can detect the target fusion results corresponding to the sensors that generate the fusion target, promptly identify sensors unsuitable for further generating the fusion target, and thus independently perform target detection on the target detection results of these sensors, avoiding the impact of these sensor target detection results on the accuracy of the fusion target, thereby improving the accuracy of target detection.

[0178] Based on the above embodiments, this application also provides a target detection device.

[0179] Reference Figure 5 , Figure 5 A structural block diagram of a target detection device 50 provided in an embodiment of this application is shown:

[0180] The acquisition module 51 is used to acquire a set of first target detection results corresponding to the first fusion target at multiple time points; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set.

[0181] The cost value module 52 is used to determine the target association cost value of the sensor set consisting of the sensors corresponding to the first target detection results based on the first target detection result set;

[0182] Sensor module 53 is used to determine the target sensor from the sensor set based on the target association value corresponding to the sensor set at the multiple times respectively;

[0183] The detection module 54 is used to independently determine the first target based on the first target detection result of the target sensor.

[0184] Optionally, the cost module includes:

[0185] The first association value submodule is used to determine the first association value between the first sensor and the second sensor other than the first sensor in the sensor set based on the first target detection result.

[0186] The target association value submodule is used to determine the target association value corresponding to the sensor set based on the first association value and the number of sensors in the sensor set.

[0187] Optionally, the sensor module includes:

[0188] The association cost change trend submodule is used to determine the association cost change trend based on the target association cost value corresponding to the sensor set at the multiple times.

[0189] The target sensor submodule is used to identify each sensor in the sensor set as a target sensor when the trend of the associated cost change is greater than or equal to a first threshold.

[0190] Optionally, the target association value includes the first association value between the first sensor and a second sensor other than the first sensor in the sensor set, and the sensor module includes:

[0191] The first quantity submodule is used to determine the first quantity corresponding to the first associated cost value of the first sensor that is less than or equal to the second threshold.

[0192] A target sensor determination submodule is used to determine the first sensor as the target sensor when the first quantity is greater than or equal to the second quantity; wherein the second quantity is obtained by subtracting the number of sensors in the sensor set from a preset quantity;

[0193] The fusion target update submodule is used to reconstruct the sensor set based on the first sensor when the first number is less than or equal to the third number, and to fuse the first target detection results corresponding to each sensor in the reconstructed sensor set to update the first fusion target; wherein the third number is less than the second number.

[0194] Optionally, the first associated cost submodule includes:

[0195] The similarity unit is used to determine the target Mahalanobis distance and target state similarity based on the first target detection result of the first sensor and the first target detection result of the second sensor;

[0196] The first association value calculation unit is used to perform a weighted summation of the target Mahalanobis distance and the target state similarity to obtain the first association value.

[0197] Optionally, the first associated cost submodule includes:

[0198] The first target velocity parameter unit is used to determine the first target velocity parameter based on the first target detection result of the first sensor;

[0199] The second target velocity parameter unit is used to determine the second target velocity parameter based on the first target detection result of the second sensor;

[0200] The first associated cost calculation unit is used to determine the first associated cost based on the difference between the first target speed parameter and the second target speed parameter.

[0201] Optionally, the device further includes:

[0202] The second fusion target module is used to acquire a second fusion target within a preset range of the first fusion target when the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than a preset distance.

[0203] The second association value module is used to determine the second association value between the third sensor of the first sensor type in the sensor set and the fourth sensor of the second sensor type in the sensor set based on the first target detection result.

[0204] The third association value module is used to determine the third association value between the third sensor and the fourth sensor based on the first target detection result and the second target detection result corresponding to the second fused target.

[0205] The third fusion target module is used to establish a fusion association relationship between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor when the third association cost is less than the second association cost, so as to fuse the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor to obtain the third fusion target.

[0206] In summary, this application provides a target detection device, comprising: acquiring a set of first target detection results corresponding to a first fusion target at multiple time points; wherein the first fusion target is obtained by fusing each first target detection result in the set of first target detection results; determining the target association value of a sensor set consisting of sensors corresponding to the first target detection results based on the set of first target detection results; determining a target sensor from the sensor set based on the target association value corresponding to the sensor set at multiple time points; and independently determining a first target through the first target detection results of the target sensor. This device can detect the target fusion results corresponding to the sensors that generate the fusion target, promptly identify sensors unsuitable for further generating the fusion target, and thus independently perform target detection on the target detection results of these sensors, avoiding the impact of these sensor target detection results on the accuracy of the fusion target, thereby improving the accuracy of target detection.

[0207] This application also provides a vehicle controller, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor performs the target detection method described above.

[0208] This application also provides a readable storage medium, which, when the instructions in the readable storage medium are executed by the processor of the vehicle controller, enables the vehicle controller to execute the above-described target detection method.

[0209] This application also provides an electronic device, which includes a processor and a memory. The memory stores a program or instructions that can run on the processor, and when the program or instructions are executed by the processor, the above-described target detection method is implemented.

[0210] This application also provides a vehicle including the aforementioned vehicle controller.

[0211] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing device embodiments, and will not be repeated here.

[0212] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

[0213] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A target detection method, characterized in that, The method includes: Obtain a set of first target detection results corresponding to the first fusion target at multiple time points; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set; Based on the first target detection result set, determine the target association value of the sensor set consisting of the sensors corresponding to the first target detection results respectively; Based on the target association value corresponding to the sensor set at the multiple times, the target sensor is determined from the sensor set; The first target is independently determined based on the first target detection result of the target sensor.

2. The method according to claim 1, characterized in that, The step of determining the target association value of the sensor set consisting of the sensors corresponding to the first target detection results based on the first target detection result set includes: Based on the first target detection result, determine the first association value between the first sensor and the second sensor other than the first sensor in the sensor set; Based on the first association generation value and the number of sensors in the sensor set, the target association generation value corresponding to the sensor set is determined.

3. The method according to claim 1, characterized in that, The step of determining the target sensor from the sensor set based on the target association value corresponding to the sensor set at the multiple times includes: Based on the target association cost value corresponding to the sensor set at the multiple times, the trend of association cost change is determined; If the trend of the associated cost change is greater than or equal to a first threshold, each sensor in the sensor set is identified as a target sensor.

4. The method according to claim 1, characterized in that, The target association value includes the first association value between the first sensor and a second sensor other than the first sensor in the sensor set. The step of determining the target sensor from the sensor set based on the target association values ​​corresponding to the sensor set at the plurality of times includes: Determine the first quantity corresponding to the first associated cost value of the first sensor that is less than or equal to the second threshold; If the first quantity is greater than or equal to the second quantity, the first sensor is determined as the target sensor; wherein, the second quantity is obtained by subtracting the number of sensors in the sensor set from a preset quantity; When the first number is less than or equal to the third number, the sensor set is reconstructed based on the first sensor, and the first target detection results corresponding to each sensor in the reconstructed sensor set are fused to update the first fused target; wherein the third number is less than the second number.

5. The method according to claim 2, characterized in that, The step of determining the first association value between the first sensor and a second sensor other than the first sensor in the sensor set based on the first target detection result includes: Based on the first target detection result of the first sensor and the first target detection result of the second sensor, the target Mahalanobis distance and target state similarity are determined; The first association value is obtained by weighted summation of the target Mahalanobis distance and the target state similarity.

6. The method according to claim 2, characterized in that, The step of determining the first association value between the first sensor and a second sensor other than the first sensor in the sensor set based on the first target detection result includes: The velocity parameters of the first target are determined based on the first target detection result of the first sensor. The velocity parameters of the second target are determined based on the first target detection result of the second sensor; The first associated cost value is determined based on the difference between the first target velocity parameter and the second target velocity parameter.

7. The method according to claim 1, characterized in that, The method further includes: If the first fusion target belongs to a preset target type and the target distance of the first fusion target is less than a preset distance, then a second fusion target within a preset range of the first fusion target is obtained; Based on the first target detection result, determine the second association value between the third sensor of the first sensor type in the sensor set and the fourth sensor of the second sensor type in the sensor set; Based on the first target detection result and the second target detection result corresponding to the second fused target, a third association generation value between the third sensor and the fourth sensor is determined; If the third association cost is less than the second association cost, a fusion association relationship is established between the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor, so as to fuse the first target detection result corresponding to the third sensor and the second target detection result corresponding to the fourth sensor to obtain the third fused target.

8. A target detection device, characterized in that, The device includes: The acquisition module is used to acquire a set of first target detection results corresponding to the first fusion target at multiple time points; wherein, the first fusion target is obtained by fusing each first target detection result in the first target detection result set. The cost value module is used to determine the target association cost value of the sensor set consisting of the sensors corresponding to the first target detection results, based on the first target detection result set. The sensor module is used to determine the target sensor from the sensor set based on the target association value corresponding to the sensor set at the multiple times. The detection module is used to independently determine the first target based on the first target detection result of the target sensor.

9. A vehicle controller, characterized in that, The vehicle controller includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the target detection method as described in any one of claims 1 to 7.

10. A vehicle, characterized in that, Includes the vehicle controller as described in claim 9.

Images