Target detection method, device, electronic device and storage medium

Abstract

The embodiments of the present invention provide a target detection method, device, electronic device and storage medium. The method includes: determining multiple target objects corresponding to a target detection area, wherein the target objects are obtained by performing preliminary target detection on the target detection area through radar; clustering the multiple target objects to be clustered to obtain at least two groups to be reviewed, each group to be reviewed corresponds to a cluster area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same shooting device and the size of the target objects in the image can be no less than the reference size when photographed simultaneously; calling the shooting device to perform target review operations on the target objects in the cluster areas corresponding to each group to be reviewed. The technical solution of the present application not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process, thereby extending the service life of the shooting equipment.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a target detection method, device, electronic equipment and storage medium. Background Art

[0002] Nowadays, the linkage between radar and ball camera has been widely used to inspect the detection area and track the target. This method can solve the problem of timeliness of target discovery, and realize full-area visual tracking of the target through ball camera monitoring.

[0003] In the scheme of implementing the linkage between radar and ball camera, due to the limitation of radar technology, the current target detection radar generally has the problem of inaccurate target recognition, and it is difficult to filter dynamic non-living targets such as trees and waves. Especially in some scenes, the false alarm rate of targets reported by radar is very high. Therefore, how to reduce these false alarms and improve the accuracy of system business data has become a difficult problem in current practical applications. Summary of the invention

[0004] The present invention provides a target detection method, device, electronic device and storage medium to solve the problem that there are too many false alarms in radar recognition results, resulting in the reception of a large amount of erroneous detection data.

[0005] In a first aspect, an embodiment of the present invention provides a target detection method, the method comprising:

[0006] Determine a plurality of target objects corresponding to the target detection area, wherein the target objects are obtained by preliminary target detection of the target detection area by radar;

[0007] Clustering the multiple target objects to be clustered to obtain at least two groups to be reviewed, each of the groups to be reviewed corresponds to a cluster area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same photographing device and the size of the target objects in the image can be no less than the reference size when photographed simultaneously;

[0008] The shooting device is called to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed.

[0009] In a second aspect, an embodiment of the present invention further provides a target detection device, the device comprising:

[0010] A determination module, used to determine a plurality of target objects corresponding to a target detection area, wherein the target objects are obtained by preliminary target detection of the target detection area by a radar;

[0011] A clustering module, used for clustering a plurality of target objects to be clustered to obtain at least two to-be-reviewed groups, each of which corresponds to a clustering area, and the target objects in the same to-be-reviewed group can be photographed simultaneously by the same photographing device and when photographed simultaneously, the size of the target objects in the image can be no less than a reference size;

[0012] The review module is used to call the shooting device to perform target review operations on the target objects in the cluster areas corresponding to each of the groups to be reviewed.

[0013] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute any target detection method in the above embodiments.

[0017] In a fourth aspect, an embodiment of the present invention further provides a computer-readable medium, which stores computer instructions, and the computer instructions are used to enable a processor to implement any target detection method in the above embodiments when executed.

[0018] The technical solution of the embodiment of the present invention is to cluster multiple target objects obtained by the radar in the preliminary detection of the target detection area, and group the many initially detected target objects by clustering, which can ensure that the size of the target objects in the same group is not less than the reference size when they are photographed by the same shooting device at the same time, and ensure that the photographed target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation. Moreover, by clustering multiple initially detected target objects, calling the shooting device to perform target review operations on the target objects in the clustering area corresponding to each group to be reviewed, it is possible to complete the review of multiple target objects at the same time when shooting each clustering area, reducing the number of areas for secondary review, avoiding the situation where the shooting device needs to be scheduled for review operations as many times as the number of target objects reported by the radar, and reducing the frequency of scheduling the shooting device for review operations, which is conducive to improving the service life of the equipment. This solution not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process, and extends the service life of the shooting equipment.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0021] Figure 1 It is a schematic diagram of a target detection method provided by an embodiment of the present invention;

[0022] Figure 2 is a flow chart of another target detection method provided by an embodiment of the present invention;

[0023] Figure 3 It is a schematic diagram of a range of reference distances of a target object relative to a group to be reviewed provided by an embodiment of the present invention;

[0024] Figure 4 It is a flowchart of another target detection method provided by an embodiment of the present invention;

[0025] Figure 5 is a schematic diagram of the structure of a target detection device provided by an embodiment of the present invention;

[0026] Figure 6 It is a structural schematic diagram of an electronic device for implementing a target detection method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0028] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0029] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0030] It should be noted that the concepts such as "first" and "second" mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0031] It should be noted that the modifications of "one" and "plurality" mentioned in the present invention are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0032] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes, and are not used to limit the scope of these messages or information.

[0033] Figure 1 A flow chart of a target detection method provided in an embodiment of the present invention. The embodiment of the present invention is applicable to the situation where a false positive detection is performed on a plurality of target objects reported when a radar reports a certain number of target objects. The target detection method can be executed by a target detection device, which can be implemented in the form of software and / or hardware, and optionally, can be implemented by an electronic device, which can be a mobile terminal, a PC or a server, etc.

[0034] like Figure 1 As shown, the target detection method provided in this embodiment may include the following process:

[0035] S110, determining a plurality of target objects corresponding to the target detection area, where the target objects are obtained by performing preliminary target detection on the target detection area through a radar.

[0036] Among them, the target detection area can be an area where radar is required for tracking and detection. The target detection area can specifically refer to the radar detection area covered and scanned by the radar used for tracking and detection. Optionally, the target detection area can refer to an area where the target object is prohibited from entering. It usually requires intensive monitoring by the monitoring equipment. Once it is determined that the target object has entered the target detection area, an alarm can be triggered. Among them, the radar can be a device that uses electromagnetic waves to detect targets. It has the characteristics of working around the clock. It emits electromagnetic waves to illuminate the target object and receives its echo, thereby obtaining information such as the distance from the target object to the electromagnetic wave emission point, the rate of change of distance (radial velocity), azimuth, and altitude.

[0037] Further optionally, the object category of the target object that is prohibited from entering the target detection area can be pre-set, that is, the alarm is triggered only when a target object enters the target detection area and the category of the target object belongs to the pre-set object category. The category of the target object that is prohibited from entering the target detection area can be set according to the detection requirements, for example, it can be a person, vehicle, etc.

[0038] As an optional but non-limiting implementation, determining multiple target objects corresponding to the target detection area includes the following steps A1-A2:

[0039] Step A1: Continuously perform target detection in the target detection area by deploying associated radars in the target detection area.

[0040] Radars associated with the target detection area are pre-deployed in the target detection area. Each radar will continuously scan and detect the target detection area, and each radar will continuously send signals and receive echoes to detect whether there is a target object in the target detection area. In addition, in order to achieve effective detection of the target, the radar can be deployed at a preset height and preset position to ensure that the detection range and accuracy of the radar achieve the expected effect, so that it can fully and accurately detect the target detection area.

[0041] Step A2: When a target is detected to appear in the target detection area, multiple target objects corresponding to the target detection area and position attribute information of the target objects are determined, and the position attribute information of the target objects includes at least one of speed, distance and azimuth angle.

[0042] When the radar detects a target in the target detection area, it will further determine multiple target objects in the target detection area. At the same time, the position attribute information of multiple target objects will also be obtained. The position attribute information may include speed (i.e., the moving speed v of the target), distance (i.e., the distance d between the target and the reference point associated with the radar, the reference point can be the radar's own position point), and the angle of the target relative to the radar (such as azimuth, pitch angle, etc.), where the azimuth refers to the angle in the horizontal direction and the pitch angle refers to the angle in the vertical direction.

[0043] It should be noted that there may be false positives among the multiple target objects reported by the radar in the target detection area. For example, a slight shaking of leaves or natural phenomena such as undulating waves may be mistakenly identified as a target object by the radar and reported.

[0044] S120. Cluster the multiple target objects to be clustered to obtain at least two groups to be reviewed, each group to be reviewed corresponds to a clustering area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same photographing device and the size of the target objects in the image can be no less than a reference size when photographed simultaneously.

[0045] Limited by the radar technology itself, radar generally has inaccurate target recognition when performing target detection, and cannot filter out dynamic non-living targets such as trees and waves. In some scenarios, more than 90% of the target objects reported by the radar are false alarms. To this end, this solution can choose to review and judge multiple target objects reported by the radar to determine which of the multiple target objects are correct and which are wrong. However, because the speed and position of the target objects are inconsistent, if the reporting order of the reported multiple target objects is used for review and judgment in sequence, some target objects may have left the target detection area before the judgment is completed, resulting in the inability to review and judge some target objects; moreover, according to the above solution, the number of target objects reported by the radar is the number of times the shooting equipment used for review needs to be dispatched, and high-frequency scheduling will affect the service life of the shooting equipment.

[0046] Based on the above situation, this scheme will cluster multiple target objects to be clustered, and assign multiple target objects to be clustered that can be photographed simultaneously by the same shooting device and whose size in the image can be no less than the reference size when photographed at the same time to the same group to be reviewed. By grouping the many target objects that have been preliminarily detected through clustering, it can be ensured that the size of the target objects in the same group when photographed simultaneously by the same shooting device in the image is no less than the reference size, thereby ensuring that the photographed target objects are clear and distinguishable, and improving the accuracy of the shooting device when performing the review operation.

[0047] Exemplarily, there are multiple target objects that need to be reviewed, and the multiple target objects are divided into several groups to be reviewed by a clustering method, each group to be reviewed includes at least one target object, and at least one group to be reviewed includes at least one target object. Each group to be reviewed corresponds to a specific clustering area. In addition, the target objects in the same group to be reviewed can be photographed simultaneously by the same shooting device, and the sizes of the multiple target objects photographed at the same time in the image can meet certain size requirements and will not be too small. Such clustering results help ensure that the captured target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation.

[0048] Optionally, the shooting device used to review the target object in this solution can be a device capable of identifying preset types of living things / vehicle characteristics, such as a camera with intelligent capabilities to identify preset types of living things / vehicles, such as a device with infrared that can perform infrared recognition; for example, this solution can be implemented using a high-definition pan-tilt camera with intelligent recognition capabilities.

[0049] As an optional but non-limiting implementation, each group to be reviewed corresponds to a cluster area, and the distance between target objects in the same group to be reviewed is no greater than a preset distance, so that adjacent and close target objects can be allocated to the same group to be reviewed.

[0050] As an optional but non-limiting implementation method, for each of the at least two groups to be reviewed, the target distance of the first target object in the group to be reviewed is greater than the reference distance corresponding to all the second target objects in the group to be reviewed, and the second target objects are the remaining target objects in the group to be reviewed except the first target object; the target distance of the first target object in the group to be reviewed is determined based on the relative distance between the first target object in the group to be reviewed and the photographing device that performs the target review operation; the reference distance corresponding to all the second target objects in the group to be reviewed is the relative distance that must be satisfied between the first target object in the group to be reviewed and the photographing device when the photographing device can simultaneously photograph all the second target objects in the group to be reviewed and meets the reference photographing condition; the reference photographing condition is used to indicate that the size of the target object captured by the photographing device in the image is not less than the reference size.

[0051] The target distance corresponding to the first target object in the group to be reviewed is determined by calculating the actual relative distance between the first target object in the group to be reviewed and the shooting device responsible for performing the target review operation. The reference distance corresponding to all the second target objects in the group to be reviewed is the relative distance between the first target object in the group to be reviewed and the shooting device when the shooting device is able to shoot all the second target objects in the group to be reviewed at the same time and the reference shooting condition is met. The reference shooting condition means that the size of the target object captured by the shooting device in the image cannot be less than the preset size.

[0052] S130, calling the shooting device to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed, respectively, where the target review operations are used to identify and determine whether the target objects in the cluster areas corresponding to the groups to be reviewed are non-living targets that are misdetected.

[0053] For each group to be reviewed, each group to be reviewed includes one or more target objects and there is at least one group to be reviewed that includes at least one target object. The shooting device is called to perform target review operations on multiple target objects in the cluster area corresponding to each group to be reviewed. The purpose of the target review operation is to identify and determine whether each target object in the cluster area corresponding to these groups to be reviewed is a non-living target that is detected by mistake. In other words, by using the shooting device to perform target review, it is determined whether the previously detected target object is a true and accurate detection result or an erroneous detection caused by some reason, and whether these target objects are non-living objects (such as static objects, objects without life features, etc.).

[0054] Optionally, when calling the photographing device to perform target review operations on the target objects in the cluster areas corresponding to each group to be reviewed, the priority order of the review operations on the cluster areas corresponding to each group to be reviewed is determined. Because the target objects included in the cluster areas corresponding to different groups to be reviewed have different importance and urgency, establishing priorities allows the photographing device to focus on key, urgent or high-risk target objects first, thereby improving the review efficiency and specificity; then, according to the priority order of the review operations on the cluster areas corresponding to each group to be reviewed, the photographing device is called to perform target review operations on the target objects in the cluster areas corresponding to each group to be reviewed in sequence.

[0055] Optionally, if only one to-be-reviewed group is obtained by clustering a plurality of target objects to be clustered, the photographing device is called to directly perform a target review operation on the target objects in the clustering area corresponding to the to-be-reviewed group.

[0056] Optionally, after determining the groups to be reviewed that contain target objects, the photographing device can be actively started or used to perform a targeted target review operation on the target objects in the cluster area corresponding to each group to be reviewed. The photographing device will perform detailed photography and data collection on the target objects in each cluster area in accordance with the set rules and procedures to complete the target review operation and ensure accurate judgment and analysis of the target objects.

[0057] Considering the situation where there are multiple target objects, and given that suspicious targets have differences in moving direction, speed, and distance from the radar, the multiple target objects reported by the radar can be clustered and classified into several pending review groups. For example, if the radar reports 8 target objects, of which 3 are in one area, 4 are in another area, and 1 is in another area, then the review of the multiple target objects can be divided into 3 times, and the target review operation of the multiple target objects can be completed after 3 reviews of the 3 and pending review groups.

[0058] For the above scheme, by clustering multiple initially detected target objects and calling the shooting equipment to perform target review operations on the target objects in the clustering areas corresponding to each group to be reviewed, it is possible to complete the review of multiple target objects at the same time when shooting each clustering area, reducing the number of areas requiring secondary review, avoiding the situation where the shooting equipment needs to be scheduled for review operations as many times as the number of target objects reported by the radar, reducing the frequency of scheduling the shooting equipment for review operations, and thus helping to increase the service life of the equipment.

[0059] The technical solution of the embodiment of the present invention is to cluster multiple target objects obtained by the radar in the preliminary detection of the target detection area, and group the many initially detected target objects by clustering, which can ensure that the size of the target objects in the same group is not less than the reference size when they are photographed by the same shooting device at the same time, and ensure that the photographed target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation. Moreover, by clustering multiple initially detected target objects, calling the shooting device to perform target review operations on the target objects in the clustering area corresponding to each group to be reviewed, it is possible to complete the review of multiple target objects at the same time when shooting each clustering area, reducing the number of areas for secondary review, avoiding the situation where the shooting device needs to be scheduled for review operations as many times as the number of target objects reported by the radar, and reducing the frequency of scheduling the shooting device for review operations, which is conducive to improving the service life of the equipment. This solution not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process, and extends the service life of the shooting equipment.

[0060] Figure 2 A flow chart of another target detection method provided for an embodiment of the present invention. The technical scheme of this embodiment further optimizes the process of clustering multiple target objects to be clustered to obtain at least two groups to be reviewed in the aforementioned embodiment on the basis of the technical scheme of the aforementioned embodiment. This embodiment can be combined with various optional schemes in one or more of the aforementioned embodiments.

[0061] like Figure 2 As shown, the target detection method provided in this embodiment may include the following process:

[0062] S210: Determine a plurality of target objects corresponding to the target detection area, where the target objects are obtained by performing preliminary target detection on the target detection area through a radar.

[0063] S220. For the current target object to be clustered among the multiple target objects to be clustered, determine the reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed; the reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed is the relative distance that must be satisfied between the current target object to be clustered and the photographing device when the photographing device can simultaneously photograph the current target object to be clustered and the target objects in the current group to be reviewed and meets the reference photographing condition; the current group to be reviewed is the group to be reviewed into which the current target object to be clustered is to be classified in the current clustering, and the reference photographing condition is used to indicate that the size of the target object captured by the photographing device in the image is not less than the reference size.

[0064] Optionally, the current group to be reviewed is the group to be reviewed that the target objects to be clustered are to be classified into in the current clustering. The current group to be reviewed is the initial group to be reviewed created by using the selected target objects to be clustered at the beginning of clustering; or, the group to be reviewed that the current clustering is to be classified into is the group to be reviewed formed by the target objects that have completed clustering during the clustering process.

[0065] When clustering multiple target objects to be clustered, cluster analysis of each target object cannot be based solely on the relative distance between the target objects, because such clustering may cause some target objects captured by the shooting device to have a relatively small imaging size in the image, resulting in the shooting device having difficulty in judging and identifying the target objects when performing target review operations because the target objects are relatively small.

[0066] To this end, this solution considers that if you want the shooting device to simultaneously shoot the current target objects to be clustered and the target objects that have been divided into the current group to be reviewed, and to ensure that when multiple target objects are shot at the same time, the size of each target object in the image is not less than the reference size, then the relative distance between the target object and the shooting device needs to be satisfied. Only by ensuring that the relative distance between the current target objects to be clustered and the shooting device meets the requirements, can it be guaranteed that the current target objects to be clustered and the target objects in the current group to be reviewed within the range of this reference distance are shot at the same time, and after meeting certain shooting requirements, the size of the current target objects to be clustered and the target objects that have been divided into the current group to be reviewed in the image is not less than the reference size, and then because the shot target objects have a sufficient size in the image, the shooting device can better perform the target review operation.

[0067] The above method is helpful to improve the accuracy and reliability of target review, and avoid the situation where the target objects captured by the shooting device are too small in the image and difficult to identify or judge due to the target objects with inappropriate shooting distances being classified into the current group to be reviewed; at the same time, the clear reference distance and shooting conditions also provide clearer standards and basis for clustering operations, making the clustering results more accurate and effective, and can better distinguish target objects of different clusters, reduce the possibility of misjudgment and missed judgment, thereby more accurately identifying the real target objects and reducing the probability of misdetecting non-living targets.

[0068] As an optional but non-limiting implementation method, determining the reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed includes the following steps B1-B3:

[0069] Step B1, determining a reference focal length corresponding to the current target object to be clustered, where the reference focal length is the maximum focal length that can be adjusted based on the ability of the shooting device to simultaneously shoot the current target object to be clustered and the target objects in the group to be reviewed.

[0070] Step B2: determining a reference size corresponding to the current target object to be clustered, where the reference size is the minimum acceptable size that the target object should meet in the image when the shooting device shoots the target object.

[0071] Step B3: Determine the reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed based on the reference focal length and the reference size.

[0072] Before setting a reference distance threshold, you need to first obtain the radar parameters and the camera parameter information corresponding to the camera device that is linked to the radar. The camera parameters include but are not limited to the following: the focal length f (unit: pixel) corresponding to the camera device, the image sensor size (width Ws, height Hs, unit: mm) and image resolution (width Wi, height Hi, unit: pixel) in the camera device; the radar parameters include but are not limited to the following: the relative distance R (unit: meter) between the target object and the camera device, and the detection angle of the target object relative to the radar (azimuth angle theta, pitch angle phi, unit: degree).

[0073] Assuming that each target object is a point source, the size of the target object in the image mainly depends on the distance and the focal length of the camera. The estimated size S (unit: pixel) of the target object in the image can be approximately calculated by the principle of similar triangles: Since the actual size of the target object in physical space is usually unknown, a typical or minimum detectable target object size can be used instead, although this value may need to be determined based on experiments or prior knowledge.

[0074] When setting a reference distance threshold, the target object whose distance from the camera used to perform the target verification operation is less than the reference distance threshold should be large enough in the image so that the camera can perform better judgment detection or recognition when performing the target verification operation. Assume that the minimum acceptable image size of the target object in the image is (Unit: pixel), then the adaptive reference distance threshold It can be calculated by the following formula: If the actual size of the target object in physical space is unknown, it can be set to a constant (C), which can be determined experimentally. It can be calculated by the following formula: .

[0075] On the basis of the above, see Figure 3 , for the current target objects to be clustered, the maximum focal length that can be adjusted by the camera to simultaneously capture the current target objects to be clustered and the target objects in the group to be reviewed can be determined, that is, the maximum focal length that can be set so that the camera can simultaneously observe the current target objects to be clustered and the target objects in the group to be reviewed is calculated. Next, determine the minimum acceptable size that the target object should meet in the image when the camera captures the target object, as the reference size, and combine the calculation formula of the reference distance threshold to obtain the reference distance threshold corresponding to the current target object to be clustered relative to the current group to be reviewed. .

[0076] As an optional but non-limiting implementation method, determining the reference focal length corresponding to the current target object to be clustered includes the following steps B11-B12:

[0077] Step B11, determine the image sensor size of the shooting device used for the target review operation on the current target objects to be clustered, and the minimum viewing angle range required for the shooting device to simultaneously cover and capture the current target objects to be clustered and the target objects in the group to be reviewed.

[0078] Step B12: determining a reference focal length corresponding to a current target object to be clustered according to the image sensor size and the minimum viewing angle range.

[0079] See also Figure 3 For a current target object q among multiple target objects to be clustered, first calculate the maximum focal length that can be adjusted by the shooting device to cover all the corresponding points of the target objects based on the area composed of q and the corresponding points of the target objects in the current group to be reviewed. The calculation example is as follows: the red point is q, the green point is the point in the current group to be reviewed, the green frame is the area composed of the original target, and the red area is the minimum coverage area of ​​the shooting device after adding q.

[0080] See also Figure 3 Based on the position of each target object and the position of the shooting device, the minimum camera coverage angle required for the shooting device to cover the target object in the to-be-reviewed group after adding point q can be calculated as , The value of , based on the longitude and latitude of A, B, C and D, can be calculated to obtain the minimum camera coverage angle required for the target object in the to-be-reviewed group, which will not be described in detail.

[0081] Furthermore, combined with the image sensor size of the shooting device used for the target review operation on the current target object to be clustered, the maximum focal length that can enable the shooting device to simultaneously observe the current target object to be clustered and the target objects in the group to be reviewed is calculated. , The width of the image sensor in the shooting device.

[0082] S230: Determine a target distance corresponding to the current target object to be clustered, where the target distance corresponding to the current target object to be clustered is a relative distance between the current target object to be clustered and a photographing device that performs a target review operation.

[0083] S240. If the target distance is less than the reference distance, the current target object to be clustered is divided into the current group to be reviewed for clustering, and the next target object to be clustered is selected from the multiple target objects to be clustered to continue clustering into the current group to be reviewed for the next clustering, or when the current target object to be clustered is the last target object to be clustered among the multiple target objects to be clustered, the current target object to be clustered is divided into the next group to be reviewed for clustering, until the division of the multiple target objects to be clustered is completed to obtain at least two groups to be reviewed, and the next group to be reviewed is the group to be reviewed to be divided into for the remaining target objects to be clustered after the clustering of the target objects to be clustered using the current group to be reviewed is completed.

[0084] Each group to be reviewed corresponds to a cluster area, and target objects in the same group to be reviewed can be photographed simultaneously by the same photographing device, and when photographed simultaneously, the size of the target objects in the image can be no smaller than the reference size.

[0085] Create an empty list C of pending review groups to store the final clustering results. Create a list U of target objects that have not yet completed clustering, and add all the corresponding points of target objects that need to be clustered to the list U. Select a corresponding point of a target object to be clustered from the list U of target objects to be clustered that have not yet completed clustering as an initial point p for clustering, and create a new current pending review group. , and add the initial point p to Then, the point p is removed from the list U corresponding to the target object that has not completed clustering.

[0086] For each target object to be clustered in the list U corresponding to the target objects that have not been clustered, take the current target object to be clustered as an example, for the corresponding point q of the current target object to be clustered in the list U corresponding to the target objects that have not been clustered, first calculate the maximum focal length that can be adjusted to cover all the corresponding points of the target objects based on the area composed of q and all the corresponding points of the target objects in the current group to be reviewed as the reference focal length, and then calculate the reference distance of the current target object to be clustered relative to the current group to be reviewed based on the reference focal length. If the target distance corresponding to the current target object to be clustered is less than the reference distance threshold of the current target object to be clustered relative to the current group to be reviewed , then add the corresponding point q of the current target object to be clustered to the current group to be reviewed and removed from the list U corresponding to the target objects that have not been clustered. Repeat this step by analogy until the list U corresponding to the target objects that have not been clustered has no more points corresponding to the current target objects to be clustered that can be added to the current group to be reviewed. middle.

[0087] When the current target object to be clustered is the last target object to be clustered among the multiple target objects to be clustered, if the list U corresponding to the target objects that have not been clustered is not empty, a point corresponding to the next target object to be clustered can be selected from the list U corresponding to the target objects that have not been clustered as a new initial point p, and a new next group to be reviewed can be created, and the above process is repeated until the list U corresponding to the target objects that have not been clustered is empty. Finally, the multiple target objects to be clustered are clustered to obtain at least two groups to be reviewed.

[0088] S250, calling the shooting device to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed, respectively, where the target review operations are used to identify and determine whether the target objects in the cluster areas corresponding to the groups to be reviewed are non-living targets that are misdetected.

[0089] The technical solution of the embodiment of the present invention is to cluster multiple target objects obtained by the radar in the target detection area, and group the many initially detected target objects by clustering, which can ensure that the size of the target objects in the same group is not less than the reference size when they are photographed by the same shooting device at the same time, and ensure that the photographed target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation. Moreover, by clustering multiple initially detected target objects, calling the shooting device to perform target review operations on the target objects in the clustering area corresponding to each group to be reviewed, it is possible to achieve the simultaneous completion of the review of multiple target objects when shooting each clustering area, reducing the number of areas for secondary review, avoiding the situation where the shooting device needs to schedule the review operation as many times as the number of target objects reported by the radar, and reducing the frequency of scheduling the review operation of the shooting device, which is conducive to improving the service life of the equipment. This solution not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process. Specifically, based on the clustering of the target area, the rotation and zoom frequency of the camera are reduced, which can effectively improve the service life of the front-end equipment such as the camera.

[0090] Figure 4 A flow chart of another target detection method provided for an embodiment of the present invention. The technical solution of this embodiment further optimizes the process of calling the shooting device in the aforementioned embodiment to perform target review operations on the target objects in the clustering areas corresponding to each of the groups to be reviewed on the basis of the technical solutions in the aforementioned embodiments. This embodiment can be combined with the various optional solutions in one or more of the aforementioned embodiments.

[0091] like Figure 4 As shown, the target detection method provided in this embodiment may include the following process:

[0092] S410, determining a plurality of target objects corresponding to the target detection area, where the target objects are obtained by performing preliminary target detection on the target detection area through a radar.

[0093] S420, clustering the multiple target objects to be clustered to obtain at least two groups to be reviewed, each group to be reviewed corresponds to a clustering area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same photographing device and the size of the target objects in the image can be no less than the reference size when photographed simultaneously.

[0094] S430: Determine a review order for at least two groups to be reviewed, where the review order is used to describe the order in which target review operations are performed on each of the at least two groups to be reviewed.

[0095] S440. Based on the review order of at least two groups to be reviewed, call the shooting device to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed in sequence. The target review operations are used to identify and determine whether each target object in the cluster areas corresponding to the groups to be reviewed is a non-living target that is misdetected.

[0096] In the current operation process, there are at least two groups to be reviewed, so it is necessary to clarify the order of review operations for these groups to be reviewed. The review order of the at least two groups to be reviewed will clearly define which group to be reviewed will be reviewed first and which one should be reviewed later when performing the target review operation on each of the multiple groups to be reviewed, and so on. The review order of the at least two groups to be reviewed can be determined based on at least one of the importance, urgency, and risk level of the target object in each group to be reviewed.

[0097] After the review order of at least two groups to be reviewed has been determined, the shooting device can be called according to the review order of at least two groups to be reviewed. The shooting device will perform target review operations on the target objects in the cluster area corresponding to each group to be reviewed one by one in order. In other words, the shooting device will first be called in order to perform a review operation on the target objects in the cluster area corresponding to the group to be reviewed that ranks first; after completion, the same operation will be performed on the group to be reviewed that ranks second, and so on, until the target review operation is completed for all groups to be reviewed.

[0098] By adopting the above method, with a clear review order, the work of the photographing equipment can be arranged more systematically, avoiding confusion and repeated operations, thereby improving the overall efficiency of the review; and the order can be determined according to the importance of different groups to be reviewed or other factors, so that the resources of the photographing equipment can be allocated more reasonably to ensure that key areas or more critical target objects are given priority. In addition, based on the various attributes of the target, the review priority of each target clustered set is comprehensively calculated, and the best scheduling strategy for the photographing equipment is made, which helps to obtain better results.

[0099] As an optional but non-limiting implementation, determining the review order of at least two to-be-reviewed groups includes the following steps C1-C3:

[0100] Step C1: for each of at least two groups to be reviewed, determine at least one reference attribute information of each target object in the group to be reviewed, the reference attribute information is used to describe the degree of attention received by each target object in the group to be reviewed when target detection is performed. The greater the degree of attention received during target detection, the higher the priority of the target object for target review operation.

[0101] Step C2: Determine the review evaluation information corresponding to the group to be reviewed based on at least one reference attribute information of each target object in the group to be reviewed, where the review evaluation information is used to describe the priority of the group to be reviewed when scheduling the shooting device to perform a target review operation on the target objects in the group to be reviewed.

[0102] Step C3: Determine the review order of at least two groups to be reviewed according to the review evaluation information corresponding to the groups to be reviewed.

[0103] For each of the at least two groups to be reviewed, at least one reference attribute information of each target object in each group to be reviewed is clarified, and these reference attribute information can reflect the degree of attention each target object receives when performing target detection. For example, the size, moving speed, and location of the target object may all be reference attributes. The greater the degree of attention received in target detection, the higher the priority of the target object in subsequent target review operations; for example, a target object that moves very fast or is particularly large in size may be considered more worthy of priority review.

[0104] Next, based on the reference attribute information of each target object in the group to be reviewed, the review evaluation information corresponding to the group to be reviewed is determined. This review evaluation information is used to describe the overall priority of the group to be reviewed when scheduling the shooting equipment to perform target review operations on the target objects in the group to be reviewed. In simple terms, it is to comprehensively consider the reference attributes of each target object in the group to derive the importance and urgency of this group in the entire review work.

[0105] Finally, the order of reviewing the at least two groups to be reviewed can be determined based on the review evaluation information corresponding to each group to be reviewed. The group to be reviewed with higher review evaluation information is more likely to be prioritized for review. In this way, the review work can be arranged reasonably, and limited resources and time can be invested in the more important groups to be reviewed that need to be reviewed in a timely manner.

[0106] Optionally, the review order is determined according to the review evaluation information corresponding to each group to be reviewed, and then one or more shooting devices for performing target review operations are controlled to review the target objects in the aggregation area corresponding to each group to be reviewed. If the review is successful, it is considered that at least one target in the aggregation area corresponding to the group to be reviewed exists, and a method similar to radar video spatial alignment can be continued to be used to further confirm which target is real and which is a false alarm. The above method can achieve accurate identification of targets in the thunderball linkage scene and prevent the generation of a large number of false alarms.

[0107] Exemplarily, determining the review order of at least two to-be-reviewed groups is considered as a multi-attribute decision analysis problem. In addition to the reference attribute information such as the speed, distance, and position of the target object, more reference attribute information can be considered, such as the size of the target object, the strength of the radar model, the shape of the target object, and the behavior mode of the target object (for example, whether it is performing an obstacle avoidance operation), etc. This can make the sorting process more detailed and comprehensive. Suppose there are n target objects , each target object has m reference attribute information , such as speed, distance, size, signal strength, etc. Target object In the reference property information The value on is recorded as For each reference attribute information , define a utility function , maps the reference attribute value x to a dimensionless score u j x ∈[0, 1] , indicating the degree of attention paid to the review priority of the reference attribute to the target object. Then, based on at least one reference attribute information of each target object in the group to be reviewed, the review evaluation information corresponding to the group to be reviewed is determined.

[0108] Optionally, determining the review evaluation information corresponding to the group to be reviewed according to at least one reference attribute information of each target object in the group to be reviewed includes the following steps:

[0109] According to at least one reference attribute information of each target object in the group to be reviewed, the review evaluation information corresponding to each target object in the group to be reviewed is determined; further, according to the review evaluation information corresponding to each target object in the group to be reviewed, the review evaluation information corresponding to the group to be reviewed is determined. The comprehensive multi-attribute score is: ,in For reference attribute information The weight of .

[0110] As an optional but non-limiting implementation, determining the review evaluation information corresponding to the group to be reviewed according to at least one reference attribute information of each target object in the group to be reviewed includes the following steps D1-D2:

[0111] Step D1. For at least one reference attribute information of each target object in the group to be reviewed, determine the reference weight information corresponding to each reference attribute information. The reference weight information is determined by dynamic adjustment based on the first reference information and the second reference information. The first reference information includes a reference vector for describing the weight adjustment factor of the reference attribute information under different environmental states and system states. The second reference information is used to describe the current environmental state and system state corresponding to the reference attribute information.

[0112] Step D2: Determine the review evaluation information corresponding to the group to be reviewed based on at least one reference attribute information of each target object in the group to be reviewed and the reference weight information corresponding to each reference attribute information.

[0113] Each reference attribute information is assigned a weight, and the weights corresponding to these reference attribute information can be dynamically adjusted according to the current situation. For example, under certain environmental conditions (such as bad weather), the weight of distance may increase because it becomes more important to respond to close target objects in sequence under such circumstances.

[0114] It is understandable that the weight is not fixed, but is a parameterized function Determined, among which is the function parameter, is a feature vector that characterizes the current environment and system state. For example, for the distance attribute The corresponding weights can be set as: ;in, is the Sigmoid function, and the linear combination The value of is mapped to the range of $(0,1)$, where Represents the dot product (or inner product) of the parameter vector and the input vector.

[0115] In a specific example, when the system detects strong winds, the weights can be adjusted dynamically. For example, the distance weight is increased to 0.5, while the speed and direction weights are reduced to 0.3 and 0.2 respectively. Such adjustments can be achieved through a parameterized function, where weather conditions (such as wind speed, precipitation, etc.) are used as part of the input feature vector z to dynamically calculate the weights of each attribute. For the distance attribute , is a parameter vector related to distance attributes, such as weight factors, environmental attenuation coefficients, etc. In a strong wind environment, we may want to increase the weight of close-range targets because strong winds may make it more difficult to track distant targets. Therefore, The parameters in may be adjusted to account for the effects of strong winds, giving greater weight to close objects.

[0116] in, Can be a vector containing multiple elements, each element corresponds to a distance attribute Weight adjustment factors for different environmental conditions. In strong wind environments, these adjustment factors may be set to larger values ​​to highlight the importance of close-range targets. The feature vector z is a feature vector that contains information about the current environment and system status. In strong wind environments, z should contain features that can characterize the effects of strong winds. For example, z can include physical quantities related to strong winds, such as wind speed, wind direction, and air density. The specific values ​​of these features can be measured by sensors and passed as inputs to the parameterized function . is the parameter vector The dot product (or inner product) of and the eigenvector z. It represents the parameter vector The degree of match between the current environment and system state z. In a strong wind environment, since z contains features that characterize the impact of strong winds, The value of will reflect the importance of the distance dimension attribute in a strong wind environment.

[0117] It should be noted that if The parameters in are set to increase the weight of close-range objects in strong wind environments, and z contains features that characterize the effects of strong winds. The value of will increase, which will increase the weight of the distance attribute, that is, the weight of the close target object becomes larger in a strong wind environment.

[0118] In another specific embodiment, at least one reference attribute information : In this embodiment, two reference attribute information are considered: Represents distance. Represents signal strength. : The weight of each reference attribute information is determined by a parameterized function, where = Corresponding distance weight, Corresponding signal strength weight. Parameter vector is specific to each reference attribute information and needs to be adjusted according to the actual application, specifically by adjusting the weight adjustment factor corresponding to the elements contained in the parameter vector. For example, and The parameter vectors represent the distance and signal strength respectively. The feature vector z is a vector that characterizes the current environment and system status, for example, it may contain information such as weather conditions, radar and ball camera performance status, etc.

[0119] For example: =[0.5,0.3,0.2]: This is an example parameter vector of the distance attribute, where the first element corresponds to the weight adjustment factor for bad weather, the second element corresponds to the weight adjustment factor for the radar performance status, and the third element corresponds to the weight adjustment factor for the ball camera performance status. =[0.4,0.6]: This is another example of a parameter vector of the signal strength attribute, used for weight calculation of the signal strength attribute. z=[1,0.8,0.9]: This is the feature vector of the current environment and system status. For example, 1 indicates that the current weather is bad, and 0.8 and 0.9 represent the performance status of the radar and ball camera respectively.

[0120] The calculation process is as follows: The first part calculates the weight of the distance attribute :First, calculate The dot product (inner product) with z is: =0.5 *1+0.3*0.8+0.2*0.9=0.86. Then, the dot product result is mapped to the range of (0,1) through the Sigmoid function: =σ( ) =σ(0.86) (Here we assume that the specific value of the Sigmoid function can be obtained by looking up a table or calculating, for example, σ(0.86)≈0.7). The second part calculates the weight of the signal strength attribute : Similarly, calculate The dot product with z: =0.4*1+0.6*0.8=0.88. Mapped to the range (0,1) through the Sigmoid function: =σ( )=σ(0.88) (assuming σ(0.88)≈0.71). Finally, we need to ensure that the sum of the weights is 1: Since dynamic adjustment may cause the sum of the weights to be unequal to 1, the weights need to be normalized. For example, the normalization is performed using the following formula: =w_1 / ( + )and = / ( + ), after normalization, ≈0.49 and ≈0.51.

[0121] Furthermore, the target object is calculated Corresponding review evaluation information assumes the target object The value of the distance attribute is =100 (unit: meter), the value of the signal strength attribute is = 8 (unit: arbitrary scale). Define the utility function Here is an example (in is the maximum value of the attribute) to convert the attribute value into a dimensionless score. Calculate the target object Review and evaluation information: . For example, For a distance of 1000 meters and a signal strength of 10, =100 / 1000=0.1, =8 / 10=0.8. Finally, =0.49*0.1+0.51* 0.8≈0.43.

[0122] As an optional but non-limiting implementation, determining the reference weight information corresponding to each reference attribute information includes the following steps E1-E2:

[0123] Step E1, determine that each reference attribute information corresponds to the current environmental state and system state, and input the current environmental state and system state into the weight adjustment model. The weight adjustment model is obtained by reinforcement learning training according to a preset reward function. The weight adjustment model records the weight adjustment factor of each reference attribute information under different environmental states and system states. The preset reward function indicates the tracking effect of the target object.

[0124] Step E2: Output reference weight information corresponding to each reference attribute information through the weight adjustment model.

[0125] System status refers to the operating status of each component or module within the system. For example, in an automated control system, the internal system status may include the speed, temperature, current, etc. of the motor; in a computer system, it may include CPU usage, memory occupancy, disk space, etc. Example: In a monitoring system in which radar and ball cameras are linked, the system status may include the following aspects: The radar detection status includes whether the radar is working, the number of targets detected, the distance, speed, direction, and other information of each target. The ball camera status refers to the position of the ball camera (indicated by PTZ parameters), viewing angle, currently tracked targets, etc. Environmental parameters refer to weather conditions, lighting, etc. as mentioned above. System performance indicators can be such as response time, tracking accuracy, error rate, etc. These indicators can reflect the overall performance of the system.

[0126] Alternatively, machine reinforcement learning methods can be used, such as modeling the entire prioritization process as a Markov decision process (MDP) through a policy network To predict the probability distribution of the next behavior, the goal is to maximize the expected reward at all times, state Including the current target attribute value, environmental status and other information, behavior To determine the target to be tracked at the next moment:

[0127] E π θ [ ∑ t =0 ∞ γ t × r t ] <none / > <none / > <mprescripts / > θ max

[0128] in represents the discount factor, reward It is an instant feedback based on the tracking effect. Network parameters Contains the weight parameters of each attribute .

[0129] In a specific example, the state : Target attribute values, including the speed of the target object , the relative distance from the target object to the camera , the size of the target object etc.; environmental conditions include visibility, weather conditions, etc., which are simplified as an environmental factor .Behavior : Determine the ID of the target to be tracked at the next moment, such as =1 means tracking target 1, =2 means tracking target 2, etc. Reward : Given according to the tracking effect, for example, a +1 reward is given for successful tracking, a -1 reward is given for lost targets, and a 0 reward is given for other situations. Policy network is a neural network, the input is the state , the output is the behavior The probability distribution of . Network parameters θ: including the weight parameters of each attribute , such as speed attribute weight , distance attribute weight , size attribute weight Etc. Discount factor (γ): used for discount when calculating cumulative rewards, usually between 0 and 1.

[0130] Among them, parameter assignment example, initial parameter assignment: =0.3, =0.5, =0.2 Discount factor: γ=0.9. Calculation process data example, initial state :The speed of target object 1 is v1=30km / h, the relative distance between target object 1 and the camera is d1=20m, and the size of target object 1 is size1="car"; the speed of target object 2 is v2=40km / h, the relative distance between target object 2 and the camera is d2=15m, and the size of target object 2 is size2="pedestrian"; environmental factors =1 (indicates normal environment); Strategy network output :Given , the policy network outputs the probability of tracking target 1 p1=0.6, and the probability of tracking target 2 p2=0.4. Select behavior : According to the probability distribution, randomly select the tracking target 1 ( =1). Perform the behavior and observe the reward : Successfully track target 1 and get reward =+1. Update status : The position, velocity and other properties of target object 1 and target object 2 change. Repeat the above steps until the termination condition is reached (such as all target objects are tracked or the maximum time step is reached). Calculate the cumulative reward: Assume that the reward is obtained in the next step , ,..., , then the cumulative reward is R= + × + × +...+ × Update network parameters θ: Use the cumulative reward R and policy gradient method to update the network parameters θ to make better decisions in future states.

[0131] By adopting the above method, the most reasonable review plan for the target is achieved through self-training and priority sorting, so as to maximize the effective review effect for all targets. With the help of intelligent technology and machine learning, the device can dynamically adjust its own weight relationship and significantly enhance its adaptability to different scenarios. Specifically, the self-training method is used to continuously optimize and improve the review strategy and process, combined with priority sorting, and accurately allocate resources and time to ensure that each target can be reviewed appropriately and in a timely manner. At the same time, based on the principles of intelligence and machine learning, the device can flexibly and autonomously adjust the weight relationship according to the changes in the actual scene, such as giving different importance weights to targets with different characteristics, so as to better adapt to various complex and changing scenarios, whether it is the number, type, distribution of targets or changes in the environment, it can respond accurately and effectively.

[0132] As an optional but non-limiting implementation, determining the reference weight information corresponding to each reference attribute information includes the following steps F1-F2:

[0133] Step F1, determining the interaction effect information between the reference attribute information according to at least one reference attribute information of each target object in the to-be-reviewed group, wherein the interaction effect information is used to describe the target tracking and recognition effect produced by the target object under the joint action of different reference attribute information;

[0134] Step F2: Determine reference weight information corresponding to each reference attribute information according to the interaction effect information between each reference attribute information.

[0135] In addition to considering the impact of multiple reference attribute information on the review priority, the interaction effect between the reference attribute information also needs to be considered. For example, for small, rapidly moving target objects (with attribute speed and attribute size), more weights may be needed to increase their review priority because such target objects may be more difficult to track and identify (the calculation of the final comprehensive score requires considering the combined impact of both speed and size).

[0136] Introduce high-order interaction terms to capture the relationship between attributes. For example, for reference attribute information , reference attribute information , introducing the interaction term:

[0137]

[0138] The function is a nonlinear interaction function that depends on the parameter The overall score is:

[0139]

[0140] Among them, the nonlinear interaction function ∅ (is a mathematical tool used to capture the nonlinear relationship between multiple attributes. In the multi-attribute decision analysis (MADA) process, when considering the impact of different attributes of the target (such as speed, distance, size, signal strength, etc.) on the priority, there is often a complex interaction between these attributes, rather than a simple linear superposition. The role of the nonlinear interaction function ∅ is to quantify this complex interaction relationship. Specifically, it can receive two or more attribute values ​​as input and output a value representing the strength of the interaction effect between these attributes. This function is usually based on certain mathematical models or algorithms, such as neural networks, polynomial regression, fuzzy logic, etc., which can express nonlinear relationships between variables. The nonlinear interaction function ∅ is used to calculate the interaction terms between attributes, which are then added to the comprehensive scoring model to more accurately evaluate the priority of the target. For example, for the two attributes of speed and size, if they are interdependent (i.e., a target with high speed and small size may require a higher priority), the nonlinear interaction function ∅ can capture this dependency and quantify it into a specific value, which is then used to adjust the final comprehensive score of the target. There are many corresponding models and algorithms to choose from, which are not listed in detail.

[0141] Weights of interaction terms It can also be adjusted dynamically through machine learning. The comprehensive scores corresponding to the above groups to be reviewed can be directly used to determine the review scores of each group to be reviewed, and then the review ranking of each group to be reviewed can be obtained. At the same time, the parameters used to sort each group to be reviewed can be built into the device before the device leaves the factory, or it can be automatically calculated by the device during operation. At the same time, the following real-time feedback mechanism can be added to the automatic calculation scheme to enable the device to calculate the optimal parameters in real time.

[0142] As an optional but non-limiting implementation, the weight allocation strategy network parameters of the weight allocation model used to allocate weights to each reference attribute information are updated online. , to achieve a continuous learning and adaptation process. First, define the current strategy to the target strategy The average KL divergence of:

[0143] D KL π * || π θ = E π * [ log π * (a|s) π θ (a|s) ]

[0144] In order to make Approach , need to be minimized , using the policy gradient method, the parameters To perform an online update: θ←θ+α E πθ [ ∇ θ logπθ ]( a | s ) A πθ ( s , a ) .

[0145] in is the learning rate, is the advantage function, which measures the behavior a relative to the current strategy The degree of dominance. s refers to the state of the environment or system.

[0146] Through the above-mentioned online updates, the policy network parameters, including the weights of various attributes, can be continuously optimized during the actual operation process, so as to continuously adapt to changes in the environment and tasks.

[0147] The technical solution of the embodiment of the present invention is to cluster multiple target objects obtained by the radar in the preliminary detection of the target detection area, and group the many initially detected target objects by clustering, which can ensure that the size of the target objects in the same group is not less than the reference size when they are photographed by the same shooting device at the same time, and ensure that the photographed target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation. Moreover, by clustering multiple initially detected target objects, calling the shooting device to perform target review operations on the target objects in the clustering area corresponding to each group to be reviewed, it is possible to complete the review of multiple target objects at the same time when shooting each clustering area, reducing the number of areas for secondary review, avoiding the situation where the shooting device needs to be scheduled for review operations as many times as the number of target objects reported by the radar, and reducing the frequency of scheduling the shooting device for review operations, which is conducive to improving the service life of the equipment. This solution not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process, and extends the service life of the shooting equipment. Based on the various attributes of the target, the review priority of each target cluster is comprehensively calculated to make the best camera scheduling strategy. Through methods based on intelligence and machine learning, the device can dynamically adjust its weight relationship to enhance scene adaptability.

[0148] Figure 5A structural schematic diagram of a target detection device provided in an embodiment of the present invention. The embodiment of the present invention is applicable to the situation where a false positive detection is performed on a plurality of target objects reported when a radar reports a certain number of target objects. The target detection device can be implemented in the form of software and / or hardware, and optionally, can be implemented by an electronic device, which can be a mobile terminal, a PC or a server, etc.

[0149] like Figure 5 As shown, the target detection device provided in this embodiment may include the following: a determination module 510, a clustering module 520 and a review module 530. Among them:

[0150] A determination module 510 is used to determine a plurality of target objects corresponding to a target detection area, wherein the target objects are obtained by performing preliminary target detection on the target detection area by radar;

[0151] A clustering module 520 is used to cluster the multiple target objects to be clustered to obtain at least two to-be-reviewed groups, each of which corresponds to a clustering area, and the target objects in the same to-be-reviewed group can be photographed simultaneously by the same photographing device and the size of the target objects in the image can be no less than the reference size when photographed simultaneously;

[0152] The review module 530 is used to call the shooting device to perform target review operations on the target objects in the cluster areas corresponding to each of the to-be-reviewed groups.

[0153] Based on the above embodiment, optionally, determining multiple target objects corresponding to the target detection area includes:

[0154] Continuously detect targets in the target detection area by deploying associated radars in the target detection area;

[0155] When a target is detected to appear in the target detection area, multiple target objects corresponding to the target detection area and position attribute information of the target objects are determined, and the position attribute information of the target objects includes at least one of speed, distance and azimuth angle.

[0156] On the basis of the above embodiment, optionally, for each of the at least two groups to be reviewed, the target distance of the first target object in the group to be reviewed is greater than the reference distance corresponding to all the second target objects in the group to be reviewed, and the second target objects are the remaining target objects in the group to be reviewed except the first target object; the target distance of the first target object in the group to be reviewed is determined based on the relative distance between the first target object in the group to be reviewed and the photographing device that performs the target review operation; the reference distance corresponding to all the second target objects in the group to be reviewed is the relative distance that must be satisfied between the first target object in the group to be reviewed and the photographing device when the photographing device can simultaneously photograph all the second target objects in the group to be reviewed and meets the reference photographing condition; the reference photographing condition is used to indicate that the size of the target object captured by the photographing device in the image is not less than the reference size.

[0157] Based on the above embodiment, optionally, clustering the multiple target objects to be clustered to obtain at least two groups to be reviewed includes:

[0158] For the current target objects to be clustered among the multiple target objects to be clustered, determine the reference distance corresponding to the current target objects to be clustered relative to the current group to be reviewed; the reference distance corresponding to the current target objects to be clustered relative to the current group to be reviewed is the relative distance that must be satisfied between the current target objects to be clustered and the shooting device when the shooting device can simultaneously shoot the current target objects to be clustered and the target objects in the current group to be reviewed and meets the reference shooting condition; the current group to be reviewed is the group to be reviewed into which the current target objects to be clustered are to be classified in the current clustering, and the reference shooting condition is used to indicate that the size of the target objects captured by the shooting device in the image is not less than the reference size;

[0159] Determine the target distance corresponding to the current target objects to be clustered, where the target distance corresponding to the current target objects to be clustered is the relative distance between the current target objects to be clustered and the shooting device that performs the target review operation;

[0160] If the target distance is less than the reference distance, the current target object to be clustered is divided into the current group to be reviewed for clustering, and the next target object to be clustered is selected from the multiple target objects to be clustered to continue to be clustered in the current group to be reviewed for the next clustering division, or when the current target object to be clustered is the last target object to be clustered participating in clustering among the multiple target objects to be clustered, the current target object to be clustered is divided into the next group to be reviewed for clustering division, until the division of the multiple target objects to be clustered is completed to obtain at least two groups to be reviewed, and the next group to be reviewed is the group to be reviewed to be divided into for the remaining target objects to be clustered after the clustering division of the target objects to be clustered using the current group to be reviewed is completed.

[0161] On the basis of the above embodiment, optionally, determining the reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed includes:

[0162] Determine a reference focal length corresponding to the current target object to be clustered, where the reference focal length is the maximum focal length that can be adjusted based on the ability of the shooting device to simultaneously shoot the current target object to be clustered and the target objects in the group to be reviewed;

[0163] Determine a reference size corresponding to the current target object to be clustered, where the reference size is the minimum acceptable size that the target object should meet in the image when the shooting device shoots the target object;

[0164] According to the reference focal length and the reference size, a reference distance corresponding to the current target object to be clustered relative to the current group to be reviewed is determined.

[0165] On the basis of the above embodiment, optionally, calling the photographing device to perform target review operations on target objects in the cluster areas corresponding to each of the to-be-reviewed groups respectively includes:

[0166] Determining a review order for the at least two groups to be reviewed, wherein the review order is used to describe a sequence in which target review operations are performed on each of the at least two groups to be reviewed;

[0167] According to the review order of the at least two groups to be reviewed, the shooting device is called to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed in sequence.

[0168] Based on the above embodiment, optionally, determining the review order of the at least two to-be-reviewed groups includes:

[0169] For each of the at least two groups to be reviewed, determining at least one reference attribute information of each target object in the group to be reviewed, wherein the reference attribute information is used to describe the degree of attention of each target object in the group to be reviewed when target detection is performed, and the greater the degree of attention, the higher the priority of the target object in the target review operation;

[0170] Determine, based on at least one reference attribute information of each target object in the group to be reviewed, review evaluation information corresponding to the group to be reviewed, wherein the review evaluation information is used to describe the priority of the group to be reviewed when scheduling a photographing device to perform a target review operation on the target objects in the group to be reviewed;

[0171] A review order for the at least two groups to be reviewed is determined according to the review evaluation information corresponding to the groups to be reviewed.

[0172] On the basis of the above embodiment, optionally, determining the review evaluation information corresponding to the group to be reviewed according to at least one reference attribute information of each target object in the group to be reviewed includes:

[0173] For at least one reference attribute information of each target object in the to-be-reviewed group, determine reference weight information corresponding to each reference attribute information, wherein the reference weight information is determined by dynamically adjusting the first reference information and the second reference information, wherein the first reference information includes a reference vector for describing a weight adjustment factor of the reference attribute information under different environmental states and system states, and the second reference information is used to describe a current environmental state and system state corresponding to the reference attribute information;

[0174] The review evaluation information corresponding to the group to be reviewed is determined according to at least one reference attribute information of each target object in the group to be reviewed and reference weight information corresponding to each reference attribute information.

[0175] Based on the above embodiment, optionally, determining that each reference attribute information corresponds to reference weight information includes:

[0176] Determine that each of the reference attribute information corresponds to a current environment state and a system state, and input the current environment state and the system state into a weight adjustment model, wherein the weight adjustment model is obtained by reinforcement learning training according to a preset reward function, and the weight adjustment model records a weight adjustment factor of each of the reference attribute information under different environment states and system states, and the preset reward function indicates the tracking effect of the target object;

[0177] The reference weight information corresponding to each reference attribute information is outputted through the weight adjustment model.

[0178] Based on the above embodiment, optionally, determining that each reference attribute information corresponds to reference weight information includes:

[0179] Determine, based on at least one reference attribute information of each target object in the to-be-reviewed group, interaction effect information between each reference attribute information, wherein the interaction effect information is used to describe target tracking and recognition effects produced by the target object under the combined effect of different reference attribute information;

[0180] According to the interaction effect information between each piece of reference attribute information, the reference weight information corresponding to each piece of reference attribute information is determined.

[0181] The technical solution provided by the embodiment of the present invention performs clustering processing on multiple target objects obtained by the radar in the preliminary detection of the target detection area, and groups the many initially detected target objects by clustering, which can ensure that the size of the target objects in the same group is not less than the reference size when they are photographed by the same shooting device at the same time, ensure that the photographed target objects are clear and distinguishable, and improve the accuracy of the shooting device when performing the review operation. Moreover, by clustering multiple initially detected target objects, calling the shooting device to perform target review operations on the target objects in the clustering area corresponding to each group to be reviewed, it is possible to achieve the simultaneous completion of the review of multiple target objects when shooting each clustering area, reducing the number of areas for secondary review, avoiding the situation where the shooting device needs to be scheduled for review operations as many times as the number of target objects reported by the radar, reducing the frequency of scheduling the shooting device for review operations, and thus helping to improve the service life of the equipment. This solution not only improves the accuracy and efficiency of target detection, but also reduces unnecessary scheduling and review operations by optimizing the process, and prolongs the service life of the shooting equipment.

[0182] The target detection device provided in the embodiment of the present invention can execute the target detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the target detection method.

[0183] It is worth noting that the various units and modules included in the above-mentioned device are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the embodiments of the present invention.

[0184] Figure 6 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Figure 6 , which shows an electronic device (eg Figure 6The terminal device in the embodiment of the present invention may include but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0185] like Figure 6 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 to a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An edit / output (I / O) interface 605 is also connected to the bus 604.

[0186] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0187] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present invention are executed.

[0188] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes, and are not used to limit the scope of these messages or information.

[0189] The electronic device provided in the embodiment of the present invention and the target detection method provided in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0190] An embodiment of the present invention provides a computer storage medium on which a computer program is stored. When the program is executed by a processor, the target detection method provided by the above embodiment is implemented.

[0191] It should be noted that the computer-readable medium of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0192] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0193] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0194] The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: determines multiple target objects corresponding to the target detection area, and the target objects are obtained by preliminary target detection of the target detection area by radar; clusters the multiple target objects to be clustered to obtain at least two groups to be reviewed, each of the groups to be reviewed corresponds to a clustering area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same shooting device and the size of the target objects in the image can be no less than the reference size when photographed at the same time; calls the shooting device to perform target review operations on the target objects in the clustering areas corresponding to each group to be reviewed.

[0195] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0196] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0197] The units involved in the embodiments of the present invention may be implemented by software or hardware. The name of a unit does not limit the unit itself in some cases. For example, the first acquisition unit may also be described as a "unit for acquiring at least two Internet Protocol addresses".

[0198] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0199] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0200] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) to form a technical solution.

[0201] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present invention. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0202] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A target detection method, characterized in that: The method comprises: Determine a plurality of target objects corresponding to the target detection area, wherein the target objects are obtained by preliminary target detection of the target detection area by radar; Clustering the multiple target objects to be clustered to obtain at least two groups to be reviewed, each of the groups to be reviewed corresponds to a cluster area, and the target objects in the same group to be reviewed can be photographed simultaneously by the same photographing device and when photographed simultaneously, the size of the target object in the image can be no less than a reference size, and the reference size is the minimum acceptable size that the target object should meet in the image when the photographing device photographs the target object; The shooting device is called to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed respectively, and the target review operations are used to identify and determine whether each target object in the cluster areas corresponding to the groups to be reviewed is a non-living target that is misdetected.

2. The method according to claim 1, characterized in that Clustering multiple target objects to be clustered obtains at least two groups to be reviewed, including: For the current target objects to be clustered among the multiple target objects to be clustered, determine the reference distance corresponding to the current target objects to be clustered relative to the current group to be reviewed; the reference distance corresponding to the current target objects to be clustered relative to the current group to be reviewed is the relative distance that must be satisfied between the current target objects to be clustered and the shooting device when the shooting device can simultaneously shoot the current target objects to be clustered and the target objects in the current group to be reviewed and meets the reference shooting condition; the current group to be reviewed is the group to be reviewed into which the current target objects to be clustered are to be classified in the current clustering, and the reference shooting condition is used to indicate that the size of the target objects captured by the shooting device in the image is not less than the reference size; Determine the target distance corresponding to the current target objects to be clustered, where the target distance corresponding to the current target objects to be clustered is the relative distance between the current target objects to be clustered and the shooting device that performs the target review operation; If the target distance is less than the reference distance, the current target object to be clustered is divided into the current group to be reviewed for clustering, and the next target object to be clustered is selected from the multiple target objects to be clustered to continue to be clustered in the current group to be reviewed for the next clustering division, or when the current target object to be clustered is the last target object to be clustered participating in clustering among the multiple target objects to be clustered, the current target object to be clustered is divided into the next group to be reviewed for clustering division, until the division of the multiple target objects to be clustered is completed to obtain at least two groups to be reviewed, and the next group to be reviewed is the group to be reviewed to be divided into for the remaining target objects to be clustered after the clustering division of the target objects to be clustered using the current group to be reviewed is completed.

3. The method according to claim 1, characterized in that Calling the photographing device to perform target review operations on target objects in the cluster areas corresponding to the groups to be reviewed respectively includes: Determining a review order for the at least two groups to be reviewed, wherein the review order is used to describe a sequence in which target review operations are performed on each of the at least two groups to be reviewed; According to the review order of the at least two groups to be reviewed, the shooting device is called to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed in sequence.

4. The method according to claim 3, characterized in that Determining a review order for the at least two groups to be reviewed includes: For each of the at least two groups to be reviewed, determining at least one reference attribute information of each target object in the group to be reviewed, wherein the reference attribute information is used to describe the degree of attention of each target object in the group to be reviewed when target detection is performed, and the greater the degree of attention, the higher the priority of the target object in the target review operation; Determine, based on at least one reference attribute information of each target object in the group to be reviewed, review evaluation information corresponding to the group to be reviewed, wherein the review evaluation information is used to describe the priority of the group to be reviewed when scheduling a photographing device to perform a target review operation on the target objects in the group to be reviewed; A review order for the at least two groups to be reviewed is determined according to the review evaluation information corresponding to the groups to be reviewed.

5. The method according to claim 4, characterized in that Determining the review evaluation information corresponding to the group to be reviewed according to at least one reference attribute information of each target object in the group to be reviewed includes: For at least one reference attribute information of each target object in the to-be-reviewed group, determine reference weight information corresponding to each reference attribute information, wherein the reference weight information is determined by dynamically adjusting the first reference information and the second reference information, wherein the first reference information includes a reference vector for describing a weight adjustment factor of the reference attribute information under different environmental states and system states, and the second reference information is used to describe a current environmental state and system state corresponding to the reference attribute information; The review evaluation information corresponding to the group to be reviewed is determined according to at least one reference attribute information of each target object in the group to be reviewed and reference weight information corresponding to each reference attribute information.

6. The method according to claim 5, characterized in that Determining reference weight information corresponding to each of the reference attribute information includes: Determine that each of the reference attribute information corresponds to a current environment state and a system state, and input the current environment state and the system state into a weight adjustment model, wherein the weight adjustment model is obtained by reinforcement learning training according to a preset reward function, and the weight adjustment model records a weight adjustment factor of each of the reference attribute information under different environment states and system states, and the preset reward function indicates the tracking effect of the target object; The reference weight information corresponding to each reference attribute information is outputted through the weight adjustment model.

7. The method according to claim 5, characterized in that Determining reference weight information corresponding to each of the reference attribute information includes: Determine, based on at least one reference attribute information of each target object in the to-be-reviewed group, interaction effect information between each reference attribute information, wherein the interaction effect information is used to describe target tracking and recognition effects produced by the target object under the combined effect of different reference attribute information; According to the interaction effect information between each piece of reference attribute information, the reference weight information corresponding to each piece of reference attribute information is determined.

8. A target detection device, characterized in that: The device comprises: A determination module, used to determine a plurality of target objects corresponding to a target detection area, wherein the target objects are obtained by preliminary target detection of the target detection area by a radar; A clustering module, used for clustering a plurality of target objects to be clustered to obtain at least two to-be-reviewed groups, each of the to-be-reviewed groups corresponding to a clustering area, and the target objects in the same to-be-reviewed group can be photographed simultaneously by the same photographing device and when photographed simultaneously, the size of the target object in the image can be no less than a reference size, and the reference size is the minimum acceptable size that the target object should meet in the image when the photographing device photographs the target object; The review module is used to call the shooting device to perform target review operations on the target objects in the cluster areas corresponding to the groups to be reviewed. The target review operation is used to identify and determine whether each target object in the cluster areas corresponding to the groups to be reviewed is a non-living target that is misdetected.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the target detection method described in any one of claims 1-7.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions are used to perform the target detection method described in any one of claims 1 to 7 when executed by a computer processor.

Images