An abnormal target filtering method, device, storage medium and program product

Abstract

This application discloses an abnormal target filtering method, apparatus, storage medium, and program product, relating to the field of image processing technology, for filtering out abnormal targets detected by radar when a reflector surface is moving. The method includes: acquiring multiple frames of detection images within a preset time period using radar, and determining multiple targets included in the multiple frames of detection images; identifying reflectors from the multiple targets, and determining the reflecting surface of the reflectors; generating multiple candidate combinations corresponding to each frame of detection image based on the reflecting surface, wherein each candidate combination includes a target located on one side of the reflecting surface and a target located on the other side of the reflecting surface in the same frame of detection image; selecting target combinations from the multiple candidate combinations corresponding to each of the multiple frames of detection images; identifying abnormal targets in the target combinations, and deleting the content information of the abnormal targets from the detection images.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an abnormal target filtering method, apparatus, storage medium and program product. Background Technology

[0002] With the development of science and technology, the types and numbers of transportation vehicles are increasing, making traffic monitoring increasingly important. Real-time traffic monitoring can better assist in the management of transportation, reduce traffic problems, and provide managers with accurate, reliable, and real-time traffic information, thus providing technical support for the realization of intelligent transportation.

[0003] In practical applications, some objects have large metallic surfaces. When other objects around these objects reflect radar waves, the waves may be reflected a second time by these objects with large metallic surfaces. Therefore, the radar image generated after receiving the echo may contain anomalous targets that are not actually existing objects. In particular, for dynamically moving objects with large metallic surfaces, such as large trucks on roads, large ships in rivers and seas, and trains on railways, the anomalous targets formed by other objects based on the reflecting surface are constantly changing because the reflecting surface is constantly moving. Traditional methods require establishing an anomalous target filtering model based on a fixed reflecting surface to filter out anomalous targets among the multiple targets detected by the radar. However, when the reflecting surface is moving, the anomalous targets also change with the reflecting surface. Traditional anomalous target filtering methods for fixed reflecting surfaces are no longer effective. To detect moving reflecting surfaces, additional detection devices or increased costs for anomalous target filtering devices are required.

[0004] Therefore, how to effectively filter out abnormal targets detected by radar when the reflector is moving is a problem that needs to be solved. Summary of the Invention

[0005] This application provides an abnormal target filtering method, apparatus, storage medium, and program product for filtering out abnormal targets detected by radar when the reflector surface is moving.

[0006] To achieve the above technical objectives, this application adopts the following technical solution:

[0007] In a first aspect, embodiments of this application provide an abnormal target filtering method, the method comprising:

[0008] The radar acquires multiple frames of detection images within a preset time period and identifies multiple targets included in the multiple frames of detection images.

[0009] Identify the reflector from multiple targets and determine the reflective surface of the reflector;

[0010] Based on the reflective surface, multiple targets in each frame of the detection image are divided into a first type of target and a second type of target. Based on the first type of target and the second type of target, multiple candidate combinations are generated for each frame of the detection image. A candidate combination includes a first type of target and a second type of target in the same frame of the detection image. The first type of target is the target located on one side of the reflective surface in the detection image, and the second type of target is the target located on the other side of the reflective surface in the detection image.

[0011] Target combinations are selected from multiple candidate combinations corresponding to multiple frames of detection images. The two targets in the target combination are symmetrical about the reflective surface in the detection image. One target in the target combination is a real target, and the other target is an abnormal target detected by radar based on the reflection of the reflective surface.

[0012] Identify anomalous targets in the target set and remove content information of anomalous targets from the detected image.

[0013] The technical solution provided in this application offers at least the following advantages: By using multiple frames of radar-detected images, reflectors and their reflecting surfaces are identified from multiple targets within the images. The detected targets are then categorized into two types: a first type located on one side of the reflecting surface and a second type located on the other side. Multiple candidate combinations are generated, each containing one first-type target and one second-type target. From these candidate combinations, a target combination containing one anomalous target and one real target is selected, allowing for the deletion of the anomalous target's information. The reflecting surface is determined by combining multiple frames of radar-detected images within a preset time period, taking into account the surface's motion. Based on this, anomalous targets are accurately identified, enabling precise filtering. Based on the radar's operating principle, anomalous targets formed by the secondary reflection of target echoes from the reflecting surface must be located on opposite sides of the corresponding real target. Therefore, classifying and combining multiple targets detected by radar based on the reflecting surface, selecting candidate combinations to identify those containing both anomalous and real targets, and then identifying the anomalous target from these combinations is a clear and uncomplicated process, easily reusable in existing equipment. Furthermore, this process only requires radar and does not require additional devices for detecting reflective surfaces, thus saving the cost of devices for filtering out abnormal targets.

[0014] In one possible implementation, selecting a target combination from multiple candidate combinations corresponding to each of the multiple detection images includes: determining multiple feature parameters corresponding to each target in each detection image based on the multiple detection images; for each detection image, selecting a reference combination from multiple candidate combinations corresponding to the detection image based on the multiple feature parameters corresponding to each target in the detection image, wherein the two targets in the reference combination are symmetrical about the reflective surface in the detection image; for each reference combination, if the ratio of the number of frames containing the reference combination in the multiple detection images to the total number of frames in the detection images within a preset time period is greater than a first preset ratio, the reference combination is determined as the target combination.

[0015] In one possible implementation, the multiple feature parameters include at least region, power, range, and velocity. Region represents the region corresponding to the target in the detection image; power represents the maximum value of multiple power peaks corresponding to the target in the detection image; range represents the distance between the target and the radar; and velocity represents the target's velocity. Based on the multiple feature parameters corresponding to each target in the detection image, a reference combination is selected from multiple candidate combinations corresponding to the detection image. This includes: for each candidate combination corresponding to the detection image, determining a symmetrical target of the first target in the candidate combination about the reflector surface, where the first target is any target in the candidate combination; if the symmetrical target corresponds to... The area of ​​the overlapping region between the first target and the area corresponding to the second target in the detection image in the candidate combination is greater than or equal to the area of ​​the second target's region. The reflection point between the first target and the second target is determined, where the second target is a target in the candidate combination other than the first target. When the reflection point is located within the region of the reflector, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, a first proportional relationship between the power of the first target and the distance to the first target, and a second proportional relationship between the power of the second target and the distance to the second target are determined. If the first proportional relationship and the second proportional relationship are different, the candidate combination is determined to be a reference combination.

[0016] In one possible implementation, the method further includes: if the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the region of the second target by a ratio less than a second preset ratio, determining that the candidate combination is not a reference combination; when the reflection point is not located within the region of the reflector, determining that the candidate combination is not a reference combination; if the difference between the velocity of the first target and the velocity of the second target is greater than a preset velocity difference, determining that the candidate combination is not a reference combination; if the first ratio relationship is the same as the second ratio relationship, determining that the candidate combination is not a reference combination.

[0017] In one possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These feature parameters include at least a region and a power. The region characterizes the region corresponding to the target in the detection image, and the power characterizes the maximum value of multiple power peaks corresponding to the target in the detection image. Determining an anomalous target in the target combination includes: determining the average power of a third target and the average power of a fourth target in the target combination. The average power characterizes the average power of the target across multiple frames of the detection image. The third target is one target in the target combination, and the fourth target is any target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target satisfy a first preset condition, the fourth target is determined to be an anomalous target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is... If the average power of the third target and the average power of the fourth target do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target are determined. The average distance is used to characterize the average value of the distances corresponding to the targets in the multi-frame detection images. If the average distance of the third target and the average distance of the fourth target meet the second preset condition, the fourth target is determined to be an abnormal target in the target combination. The second preset condition is that the average distance of the third target is greater than the average distance of the fourth target, and the difference between the average distance of the third target and the average distance of the fourth target is greater than the second preset difference. If the average distance of the third target and the average distance of the fourth target do not meet the second preset condition, the movement curve of the third target and the movement curve of the fourth target are determined based on the multi-frame detection images. If the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the fourth target is determined to be an abnormal target in the target combination.

[0018] In one possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. Determining a reflector from multiple targets and determining the reflective surface of the reflector includes: for each frame of the detection image, dividing the detection image into multiple non-overlapping image blocks; if the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one edge of the region corresponding to the fifth target is greater than a preset length, then the fifth target is marked as a reference target; if the ratio of the number of frames in which the fifth target is marked as a reference target to the total number of frames in the detection image within a preset time period is greater than or equal to a third preset ratio, then the fifth target is determined to be a reflector; the surface containing the edge with a length greater than a preset length in the region corresponding to the reflector is determined as the reflective surface of the reflector.

[0019] Secondly, this application provides an abnormal target filtering device, including a processing module, which is used for:

[0020] The radar acquires multiple frames of detection images within a preset time period and identifies multiple targets included in the multiple frames of detection images.

[0021] Identify the reflector from multiple targets and determine the reflective surface of the reflector;

[0022] Based on the reflective surface, multiple targets in each frame of the detection image are divided into a first type of target and a second type of target. Based on the first type of target and the second type of target, multiple candidate combinations are generated for each frame of the detection image. A candidate combination includes a first type of target and a second type of target in the same frame of the detection image. The first type of target is the target located on one side of the reflective surface in the detection image, and the second type of target is the target located on the other side of the reflective surface in the detection image.

[0023] Target combinations are selected from multiple candidate combinations corresponding to multiple frames of detection images. The two targets in the target combination are symmetrical about the reflective surface in the detection image. One target in the target combination is a real target, and the other target is an abnormal target detected by radar based on the reflection of the reflective surface.

[0024] Identify anomalous targets in the target set and remove content information of anomalous targets from the detected image.

[0025] In one possible implementation, the processing module is specifically used to: determine multiple feature parameters corresponding to each target in each frame of the detection image based on multiple frames of detection images; for each frame of detection image, select a reference combination from multiple candidate combinations corresponding to the detection image based on the multiple feature parameters corresponding to each target in the detection image, wherein the two targets in the reference combination are symmetrical about the reflective surface in the detection image; for each reference combination, if the number of frames in the multiple frames of detection images that contain the reference combination is greater than the ratio of the total number of frames of detection images within a preset time period to a first preset ratio, determine the reference combination as the target combination.

[0026] In one possible implementation, the multiple feature parameters include at least region, power, distance, and velocity. Region represents the region corresponding to the target in the detection image; power represents the maximum value of multiple power peaks corresponding to the target in the detection image; distance represents the distance between the target and the radar; and velocity represents the target's velocity. Specifically, the processing module is used to: for each candidate combination corresponding to the detection image, determine the symmetrical target of the first target in the candidate combination with respect to the reflector surface, where the first target is any target in the candidate combination; if the area of ​​the overlapping portion between the region corresponding to the symmetrical target in the detection image and the region corresponding to the second target in the detection image in the candidate combination is greater than or equal to the area of ​​the second target's region, determine the reflection point between the first target and the second target, where the second target is a target in the candidate combination other than the first target; when the reflection point is located within the reflector's region, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, determine a first proportional relationship between the power of the first target and the distance of the first target, and a second proportional relationship between the power of the second target and the distance of the second target; if the first proportional relationship differs from the second proportional relationship, determine that the candidate combination is a reference combination.

[0027] In one possible implementation, the processing module is further configured to: determine that the candidate combination is not a reference combination if the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the region of the second target by a ratio less than a second preset ratio; determine that the candidate combination is not a reference combination when the reflection point is not located within the region of the reflector; determine that the candidate combination is not a reference combination if the difference between the velocity of the first target and the velocity of the second target is greater than a preset velocity difference; and determine that the candidate combination is not a reference combination if the first ratio relationship is the same as the second ratio relationship.

[0028] In one possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These feature parameters include at least a region and a power. The region characterizes the region corresponding to the target in the detection image, and the power characterizes the maximum value of multiple power peaks corresponding to the target in the detection image. Specifically, the processing module is used to: determine the average power of the third target and the average power of the fourth target in the target combination. The average power characterizes the average power of the target across multiple frames of the detection image. The third target is one target in the target combination, and the fourth target is any target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target satisfy a first preset condition, the fourth target is determined to be an abnormal target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is greater than the average power of the third target. A preset difference is set; if the average power of the third target and the average power of the fourth target do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target are determined. The average distance is used to characterize the average value of the distances corresponding to the targets in the multi-frame detection images; if the average distance of the third target and the average distance of the fourth target meet the second preset condition, the fourth target is determined to be an abnormal target in the target combination. The second preset condition is that the average distance of the third target is greater than the average distance of the fourth target, and the difference between the average distance of the third target and the average distance of the fourth target is greater than the second preset difference; if the average distance of the third target and the average distance of the fourth target do not meet the second preset condition, the movement curve of the third target and the movement curve of the fourth target are determined based on the multi-frame detection images; if the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the fourth target is determined to be an abnormal target in the target combination.

[0029] In one possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. The processing module is specifically used to: for each frame of the detection image, divide the detection image to obtain multiple non-overlapping image blocks; if the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one side of the region corresponding to the fifth target is greater than a preset length, mark the fifth target as a reference target; if the number of frames in which the fifth target is marked as a reference target in multiple frames of the detection image is greater than or equal to a third preset ratio to the total number of frames of the detection image within a preset time period, determine the fifth target as a reflector; and determine the surface where the side with a length greater than a preset length in the region corresponding to the reflector is determined as the reflecting surface of the reflector.

[0030] Thirdly, this application provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any of the abnormal target filtering methods provided in the first aspect above.

[0031] Fourthly, this application provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the steps of any of the abnormal target filtering methods provided in the first aspect.

[0032] Fifthly, this application provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of any of the abnormal target filtering methods provided in the first aspect.

[0033] For a detailed description of the second to fifth aspects and their various implementations in this application, please refer to the detailed description in the first aspect and its various implementations; and for a detailed analysis of the beneficial effects of the second to fifth aspects and their various implementations in the first aspect and its various implementations, please refer to the beneficial effect analysis in the first aspect and its various implementations, which will not be repeated here.

[0034] These or other aspects of this application will become more readily apparent in the following description. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of an abnormal target formation scenario provided in an embodiment of this application;

[0036] Figure 2 This is a schematic diagram of the architecture of a detection system used in an abnormal target filtering method provided in an embodiment of this application;

[0037] Figure 3 A schematic diagram of a radar structure provided in an embodiment of this application;

[0038] Figure 4 This is a schematic diagram of the hardware composition of a computing device provided in an embodiment of this application;

[0039] Figure 5 A flowchart illustrating an abnormal target filtering method provided in an embodiment of this application;

[0040] Figure 6 This application provides an illustration of a use case for an abnormal target filtering method. Figure 1 ;

[0041] Figure 7 This application provides an illustration of a use case for an abnormal target filtering method. Figure 2 ;

[0042] Figure 8 This application provides an illustration of a use case for an abnormal target filtering method. Figure 3 ;

[0043] Figure 9 This is a schematic diagram of an abnormal target filtering device provided in an embodiment of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] It should be noted that, in the embodiments of this application, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0046] To facilitate understanding, we will first provide a brief introduction and explanation of some terms or basic concepts of technology involved in the embodiments of this application.

[0047] Radar, a transliteration of the English word "radar," is an abbreviation of "Radio Detection and Ranging." It uses radio waves to detect targets and determine their positions; therefore, radar is also known as "radio positioning." Radar is an electronic device that uses electromagnetic waves to detect targets, such as millimeter-wave radar, microwave radar, and ultra-wideband radar. Radar illuminates a target with electromagnetic waves and receives the reflected echoes, thereby obtaining information such as the distance from the target to the electromagnetic wave emission point, the rate of change of distance (radial velocity), azimuth, and altitude. The principle of radar distance measurement is that by measuring the time difference between the emitted and received electromagnetic waves, the distance between the target and the radar can be obtained. The principle of radar azimuth measurement is that the radar measures the distance and elevation angle based on the azimuth and elevation beams of the antenna, thus obtaining the target's azimuth relative to the radar.

[0048] In this application, an anomalous target refers to a false target that does not actually exist, whose radar waves are reflected by other reflective surfaces and then received by the radar. For example, such as Figure 1 As shown in (a), the radar wave reflected by the car A0 is reflected by the surface of the truck B0 and received by the radar. Based on the radar wave's transmission time, the direction of the received radar wave's echo, and the time of receipt, the radar determines that a target A exists on the other side of the truck B0. ′ 0, this target A ′ The abnormal target 0 is formed by the secondary reflection of radar waves from truck B0 onto car A0; target A does not actually exist. ′ 0.

[0049] The above is an introduction to some of the concepts involved in the embodiments of this application, which will not be repeated below.

[0050] With the development of science and technology, the types and numbers of transportation vehicles are increasing, making traffic monitoring increasingly important. Real-time traffic monitoring can better assist in the management of transportation, reduce traffic problems, and provide managers with accurate, reliable, and real-time traffic information, thus providing technical support for the realization of intelligent transportation.

[0051] In practical applications, some objects have large metallic surfaces. When other objects around these objects reflect radar waves, the waves may be reflected a second time by these objects with large metallic surfaces. Therefore, the radar image generated after receiving the echo may contain anomalous targets that are not actually existing objects. In particular, for dynamically moving objects with large metallic surfaces, such as large trucks on roads, large ships in rivers and seas, and trains on railways, the anomalous targets formed by other objects based on the reflecting surface are constantly changing because the reflecting surface is constantly moving.

[0052] For example, such as Figure 1 As shown in (a), a car A0 is driving to the left of a truck B0. The outer surface of the truck B0 forms a strong reflective surface. The radar waves reflected by the car A0 are reflected twice by this reflective surface, and the radar detection result indicates that there is an abnormal target A on the right side of the truck B0. ′ 0. When the car A0 and the truck B0 are in motion, if the speed of the car A0 is greater than the speed of the truck B0, the abnormal target detected by the radar and the states of the car A0 and the truck B0 can be compared as follows: Figure 1 As shown in (b), if the speed of the car A0 is less than the speed of the truck B0, the abnormal target detected by the radar and the states of the car A0 and the truck B0 can be as follows: Figure 1As shown in (c) in the figure. It can be seen that the reflection point of the radar wave reflected by the reflector is different depending on the motion state, and the location of the abnormal target is also different. The traditional scheme of filtering out abnormal targets according to the fixed reflector is no longer effective, resulting in false detection of targets in this case.

[0053] In traditional methods, an abnormal target filtering model needs to be established based on a fixed reflector to filter out abnormal targets among multiple targets detected by radar. However, when the reflector moves, the abnormal targets also change with the reflector. The traditional abnormal target filtering method corresponding to a fixed reflector is no longer effective. If it is necessary to detect moving reflectors, other detection devices need to be added, or the cost of the device for filtering abnormal targets needs to be increased.

[0054] Therefore, how to effectively filter out abnormal targets detected by radar when the reflector is moving is a problem that needs to be solved.

[0055] To address this issue, this application provides an abnormal target filtering method. Using multiple frames of radar-detected images, reflectors and their reflecting surfaces are identified from among the multiple targets present in the images. The detected targets are then categorized into a first type of target located on one side of the reflecting surface and a second type of target located on the other side, generating multiple candidate combinations. Each candidate combination includes one first-type target and one second-type target. From these candidate combinations, a target combination containing one abnormal target and one real target is selected, and the abnormal target's information is then deleted. By combining multiple frames of radar-detected images within a preset time period to determine the reflecting surface and taking its movement into account, abnormal targets are accurately identified based on this reflecting surface, enabling precise filtering. Based on the working principle of radar, abnormal targets formed by the secondary reflection of target echoes from the reflecting surface must be located on opposite sides of the reflecting surface along with their corresponding real targets. Therefore, classifying and combining multiple targets detected by radar based on the reflecting surface, selecting candidate combinations to determine target combinations containing both abnormal and real targets, and then identifying the abnormal target from these combinations—this processing logic is clear and uncomplicated, making it easy to reuse in existing equipment. Furthermore, this process only requires radar and does not require additional devices for detecting reflective surfaces, thus saving the cost of devices for filtering out abnormal targets.

[0056] Please refer to Figure 2 This illustrates the detection system to which the abnormal target filtering method provided in this application is applicable. For example... Figure 2 As shown, the detection system 1 includes: radar 10 and electronic equipment 20.

[0057] The radar 10 and the electronic device 20 establish a communication connection directly or indirectly. It should be understood that the connection method can be wireless, such as Bluetooth or Wi-Fi; or it can be wired, such as fiber optic, etc., without limitation. For example, the radar 10 and the electronic device 20 can be connected via a wireless local area network.

[0058] In some embodiments, the radar 10 is used to detect targets within a preset range and generate a detection image from which characteristic parameters of each target can be read. For example, the preset range may be intersections of traffic roads, docks, etc., and the radar can be used to detect vehicles within these ranges.

[0059] Specifically, such as Figure 3 As shown, radar 10 may include a radar transmitter 101, a radar receiver 102, an antenna 103, and a transceiver switch 104.

[0060] Radar transmitter 101 is a radio device that provides high-power radio frequency signals to radar 10. It is capable of generating high-power radio frequency signals, i.e., electromagnetic waves, with a modulated carrier wave. Based on the modulation method, transmitters can be divided into two categories: continuous wave transmitters and pulse transmitters. A transmitter consists of a single-stage radio frequency oscillator and a pulse modulator.

[0061] Radar receiver 102 is a device for frequency conversion, filtering, amplification, and demodulation in radar 10. Through appropriate filtering, it selects the weak high-frequency signal received by the antenna from accompanying noise and interference, and after amplification and detection, uses it for target detection, display, or other radar signal processing.

[0062] Antenna 103 is a device in radar 10 used to transmit or receive electromagnetic waves and determine their detection direction. When transmitting, it concentrates energy to radiate in the direction to be illuminated; when receiving, it receives the echo in the detection direction and distinguishes the target's azimuth and / or elevation angle.

[0063] When radar 10 transmits a signal, transceiver switch 104 connects antenna 103 to radar transmitter 101 and disconnects it from radar receiver 102 to prevent high-power transmitted signals from entering radar receiver 102 and burning out the high-frequency amplifier or mixer. When radar 10 receives a signal, transceiver switch 104 connects antenna 103 to radar receiver 102 and disconnects it from radar transmitter 101 to prevent the weak received signal from being lost due to bypassing radar transmitter 101.

[0064] In some embodiments, the electronic device 20 is used to process multiple frames of detection images obtained by the radar 10, identify abnormal targets in the detection images, and delete the content information of the abnormal targets in the detection images.

[0065] In some embodiments, the electronic device 20 may include a human-computer interaction device for displaying the detected image after filtering out abnormal targets to a user, and for receiving the user's viewing operations on the detected images after filtering out abnormal targets at different time points. For example, the human-computer interaction device may include a display, such as a liquid crystal display, an organic light-emitting diode (OLED) display, etc., and the specific type, size, and resolution of the display are not limited. As another example, the human-computer interaction device may also include a voice recognition device to receive voice commands issued by the user. Furthermore, the human-computer interaction device may also include a keyboard, touchpad, touchscreen, remote control, or handwriting device. This application does not limit the specific form of the human-computer interaction device.

[0066] In some embodiments, electronic device 20 may be a single server or a server cluster, or it may be a terminal device, such as a personal computer (PC), laptop computer, mobile device, tablet computer, etc. This application does not limit the specific form of electronic device 20.

[0067] In some embodiments, the detection system 1 may also include multiple electronic devices 20 to facilitate the simultaneous processing of abnormal target filtering on the detection images obtained by multiple radars 10.

[0068] In some embodiments, the radar 10 and the electronic device 20 can be as follows: Figure 2 The diagram shows two separate devices, or the radar 10 and the electronic device 20 can be integrated together; this application does not impose specific limitations on this.

[0069] The hardware structure of the aforementioned electronic device 20 includes Figure 4 The components included in the computing device shown. The following are examples... Figure 4 Taking the computing device shown as an example, the hardware structure of electronic device 20 will be introduced.

[0070] like Figure 4 As shown, the computing device may include a processor 301, a memory 302, a communication interface 303, and a bus 304. The processor 301, the memory 302, and the communication interface 303 can be connected via the bus 304.

[0071] Processor 301 is the control center of the computing device. It can be a single processor or a collective term for multiple processing elements. For example, processor 301 can be a general-purpose central processing unit (CPU) or other general-purpose processors. Among them, the general-purpose processor can be a microprocessor or any conventional processor.

[0072] As one embodiment, processor 301 may include one or more CPUs, for example Figure 3 CPU 0 and CPU 1 are shown in the diagram.

[0073] The memory 302 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0074] In one possible implementation, the memory 302 can exist independently of the processor 301. The memory 302 can be connected to the processor 301 via a bus 304 and is used to store instructions or program code. When the processor 301 calls and executes the instructions or program code stored in the memory 302, it can implement the model deployment method provided in the embodiments of this application.

[0075] In another possible implementation, the memory 302 can also be integrated with the processor 301.

[0076] Communication interface 303 is used for connecting the computing device to other devices via a communication network, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc. Communication interface 303 may include a receiving unit for receiving data and a transmitting unit for transmitting data.

[0077] Bus 304 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0078] It should be pointed out that, Figure 4 The structure shown does not constitute a limitation on the computing device, except Figure 4 In addition to the components shown, the computing device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0079] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0080] like Figure 5 As shown, this application embodiment provides an abnormal target filtering method, which can be executed by the aforementioned electronic device 20. The method includes the following steps:

[0081] S101. Acquire multiple frames of detection images within a preset time period using radar, and determine the multiple targets included in the multiple frames of detection images.

[0082] Radar can be used to detect targets. For example, in traffic flow prediction, radar can be used to detect vehicles on roads and thus determine the traffic volume. Another example is in detecting the number of ships at a dock, where radar can be used to easily count the number of ships docked.

[0083] In practical applications, radar transmits electromagnetic waves and receives multiple frames of echo signals. Each frame is processed to obtain a multi-frame detection image, corresponding one-to-one with the echo signals. For example, the transmitted electromagnetic waves and received echo signals can use linear frequency modulated continuous wave (LFMCW), which effectively reduces the probability of interception and interference. Since LFMCW signals have extremely long time spans, pulse compression processing cannot be performed in the time domain. The radar can first perform a Fourier transform on the multi-frame echo signals, converting them from time-domain signals to frequency-domain signals. This allows for more intuitive pulse compression and moving target indication processing of each echo signal to obtain the detection image. The detection image includes a point cloud map and a power map. A single echo signal frame contains multiple echo signals, and each echo signal can identify a point. The radar detects a point cloud from a single echo signal frame. Electronic equipment can then cluster these radar-detected point clouds into targets, resulting in a point cloud map. This point cloud map contains multiple targets identified through point cloud clustering. From the point cloud map, information such as the region corresponding to each target, its size, and its location can be obtained. Furthermore, the electronic equipment can also obtain a power map from the radar detection, from which target characteristic parameters can be read, such as the peak power of each target, the target's velocity, and the distance between the target and the radar.

[0084] For example, Figure 6 This image shows a power graph obtained from a radar detection. The darkness of the color in the power graph represents the power level, i.e., the strength of the reflected radar wave energy. The darker the color, the higher the power and the stronger the reflected radar wave energy; the lighter the color, the lower the power and the weaker the reflected radar wave energy. The horizontal axis of the power graph represents velocity, with the leftmost and rightmost edges indicating a velocity of 0. The vertical axis is located in the center of the power graph. The two regions divided by the vertical axis have opposite velocities. For example, the velocity on the left side of the vertical axis is positive, and the velocity on the right side is negative. Positive can be set to the direction away from the radar, and negative to the direction approaching the radar; conversely, positive can be set to the direction approaching the radar, and negative to the direction away from the radar. The closer the horizontal axis is to the vertical axis, the greater the velocity. The vertical axis of the power graph represents the distance to the radar. From this power graph, targets A, B, and C can be identified. Target A has the darkest color, and target C has the lightest color. The power of target A is 80 dB, target B is 50 dB, and target C is 35 dB. Assuming target A is identified as a reflector based on the point cloud map, the power graph shows that target A is moving towards the radar, while targets B and C are moving away from the radar, and their speeds are the same. The power graph also shows that the distance between target A and the radar is 33 meters, the distance between target B and the radar is 30 meters, and the distance between target C and the radar is 40 meters.

[0085] In some embodiments, the preset duration can be a period of time set by the user after the radar transmits radar wave signals. For example, the user sets the radar to transmit radar waves at a preset frequency, and receives multiple frames of echo signals within 1 second after each radar wave transmission, obtaining multiple frames of detection images. The electronic device can then determine the reflectors in the multiple frames of detection images within this 1-second period, as well as the abnormal targets formed by the reflection of the reflectors.

[0086] S102. Identify the reflector from multiple targets and determine the reflecting surface of the reflector.

[0087] After identifying multiple targets detected in the image, reflectors can be identified from these targets, and then the reflecting surfaces of the reflectors can be determined.

[0088] In some embodiments, the electronic device can determine reflectors among multiple targets based on multiple feature parameters corresponding to each target in each frame of the detection image, and then determine the reflecting surface of the reflector. The multiple feature parameters include at least a region and a power, where the region characterizes the region corresponding to the target in the detection image, and the power characterizes the maximum value of multiple power peaks corresponding to the target in the detection image.

[0089] For example, step S102 can be specifically implemented as follows: the electronic device divides the detection image to obtain multiple non-overlapping image blocks; if the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one side of the region corresponding to the fifth target is greater than a preset length, the electronic device marks the fifth target as a reference target; if the number of frames in which the fifth target is marked as a reference target in multiple frames of detection images is greater than or equal to the ratio of the total number of frames of detection images within a preset time period to a third preset ratio, the electronic device determines the fifth target as a reflector; then the electronic device determines the surface of the region corresponding to the reflector where the side with a length greater than a preset length is located as the reflecting surface of the reflector.

[0090] Optionally, the electronic device can mark the fifth target as a reference target by adding a suspected reflector identifier to the fifth target. For example, the suspected reflector identifier can be the number 1, and the electronic device will associate the fifth target with 1 and other targets not marked as reflectors with 0.

[0091] For example, assume a preset quantity of 6, a preset power of 60dB, a preset length of 2 meters, and a third preset ratio of 80%. Based on the above example, if the number of image blocks occupied by the region corresponding to target B is 7, and the length of the left boundary of the region of target B is 3 meters. If the power of target B is 70dB, target B can be marked as a reflector. If the radar detects 20 frames of detection images within a preset time period, and target B is marked as a reflector in 19 of those frames, then the ratio of the number of frames in which target B is marked as a reflector to the total number of frames in the preset time period is 95%. 95% > 80%, therefore target B can be identified as a reflector, and its left boundary is determined to be a reflective surface. If the power of target B is 55dB, then target B will not be marked as a reflector.

[0092] In this way, by dividing the detected image into multiple image blocks, if the number of image blocks occupied by the region corresponding to the fifth target is greater than a preset number, it indicates that the fifth target has a large volume, meeting the volume requirement for a reflector. If the power of the fifth target is greater than a preset power, it indicates that the energy of the echo signal reflected by the fifth target is strong enough, and the fifth target has strong reflectivity, meeting the reflectivity requirement for a reflector. If the length of one side of the fifth target is greater than a preset length, it indicates that this side is likely to be the reflective surface of the target, meeting the requirement that the target has a reflective surface. By setting a third preset ratio, the credibility of the target being a reflector over a period of time can be determined. Through the above conditions, the electronic equipment can accurately determine the reflective surface of the reflector among the targets detected by the radar, facilitating the subsequent filtering of abnormal targets caused by reflection from the reflective surface.

[0093] For example, step S102 can be specifically implemented as follows: if a target among multiple targets satisfies any two of the three preset conditions, the electronic device marks the target as a reference target. Furthermore, if the ratio of the number of frames in which the target is marked as a reference target to the total number of frames in the detected images within a preset time period is greater than or equal to a third preset ratio, the electronic device determines that the target is a reflector; and then the electronic device determines the reflecting surface of the reflector. The three preset conditions are: the electronic device divides the detected image into multiple non-overlapping image blocks; the number of image blocks occupied by the region of the fifth target in the detected image is greater than a preset number; the power of the fifth target is greater than a preset power; and the length of at least one side of the region corresponding to the fifth target is greater than a preset length.

[0094] For example, suppose the preset quantity is 6, the preset power is 60dB, the preset length is 2 meters, and the third preset ratio is 80%. If the number of image blocks occupied by the detected target combination area is 7, the length of the left boundary of the target area is 3 meters, and the target power is 40dB, satisfying two preset conditions, then the electronic device can mark the target as a reflector. If the radar detects 20 frames of detection images within a preset time period, and the target is marked as a reflector in 19 of those frames, then it can be determined that the ratio of the number of frames in which the target is marked as a reflector to the total number of frames in the preset time period is 95%. Since 95% > 80%, the target can be identified as a reflector, and the left boundary of the target can be identified as a reflective surface.

[0095] In this way, by setting conditions such as area, power, and size, electronic devices can screen out targets with a high probability of being reflectors as reflectors, determine the reflecting surface of the reflector, and facilitate the subsequent filtering out of abnormal targets caused by reflection from the reflecting surface.

[0096] In some embodiments, if the number of frames in which the fifth target is marked as a reference target in a multi-frame detection image within a first time period is greater than or equal to a third preset ratio to the total number of frames in the detection image within the first time period, the electronic device determines that the fifth target is a reflector in the first time period and adds a reflector identifier to the fifth target. If the number of frames in which the fifth target is marked as a reference target in a multi-frame detection image within a second time period is less than a fourth preset ratio to the total number of frames in the detection image within a preset duration in the second time period, the electronic device determines that the fifth target is not a reflector in the second time period and deletes the added reflector identifier for the fifth target. The first time period is earlier than the second time period.

[0097] Optionally, reflector identification can take many forms. For example, reflector identification can be in numerical form; the electronic device adds the number 2 to the fifth target to determine that the fifth target is a reflector.

[0098] For example, assuming the third preset ratio is 80% and the fourth preset ratio is 50%, if 25 detection images are detected between 10:21:00 and 10:21:01, and target A is marked as a reflective object in 21 of these images, then the electronic device determines that target A is a reflective object between 10:21:00 and 10:21:01 and adds a reflective object identifier to target A. If 25 detection images are detected between 10:21:01 and 10:21:02, and target A is marked as a reflective object in 10 of these images, then the electronic device determines that target A is not a reflective object between 10:21:01 and 10:21:02 and deletes the reflective object identifier added to target A between 10:21:00 and 10:21:01.

[0099] In this way, by setting the fourth preset ratio, the credibility of the target not being a reflector within a certain period of time can be determined. Thus, electronic equipment can determine in real time whether the target is a reflector, is more adaptable to the motion state of the reflector, and can more accurately identify abnormal targets detected by radar in subsequent processes.

[0100] In some embodiments, after identifying a reflector in the current frame detection image, the electronic device performs planar detection on the detected point cloud corresponding to the detection image to obtain the reflecting surface of the reflector. For example, the Hough transform can be used to perform planar detection on the detected point cloud to obtain the reflecting surface of the reflector.

[0101] S103. Based on the reflective surface, the multiple targets in each frame of the detection image are divided into a first type of target and a second type of target, and multiple candidate combinations corresponding to each frame of the detection image are generated based on the first type of target and the second type of target.

[0102] One candidate combination includes a first type of target and a second type of target. The first type of target is the target located on one side of the reflective surface in the detected image, and the second type of target is the target located on the other side of the reflective surface in the detected image.

[0103] For example, a frame of detected image includes a reflector X, target A, target B, target C, target D, target E, and target F. The reflector X has a reflective surface. Target A and target B are located to the left of the reflective surface of reflector X, and targets C, D, E, and F are all located to the right of the reflective surface of reflector X. Then, the candidate combinations determined by the electronic device are: candidate combination 1 (target A and target C), candidate combination 2 (target A and target D), candidate combination 3 (target A and target E), candidate combination 4 (target A and target F), candidate combination 5 (target B and target C), candidate combination 6 (target B and target D), candidate combination 7 (target B and target E), and candidate combination 8 (target B and target F).

[0104] In some embodiments, if multiple reflective surfaces exist in the detected image, the electronic device generates multiple candidate combinations for each reflective surface.

[0105] For example, a frame of detected image includes target A, target B, target C, target D, target E, and target F, where target A has a reflective surface x1 and target B has a reflective surface x2. With respect to reflective surface x1, targets B and C are located to the left of reflective surface x1, while targets D, E, and F are all located to the right of reflective surface x1. Therefore, the electronic device generates the following multiple candidate combinations for reflective surface x1: candidate combination 11 (target B and target D), candidate combination 12 (target B and target E), candidate combination 13 (target B and target F), candidate combination 14 (target C and target D), candidate combination 15 (target C and target E), and candidate combination 16 (target C and target F). For reflective surface x2, targets A, C, and D are located to the left of reflective surface x2, while targets E and F are both located to the right. Therefore, the electronic device generates the following candidate combinations for reflective surface x2: Candidate Combination 21 (Target A and Target E), Candidate Combination 22 (Target A and Target F), Candidate Combination 23 (Target C and Target E), Candidate Combination 24 (Target C and Target F), Candidate Combination 26 (Target D and Target E), and Candidate Combination 27 (Target D and Target F). In the subsequent process of identifying anomalous targets, the anomalous targets are determined based on the candidate combinations generated for different reflective surfaces.

[0106] In this way, the electronic device generates multiple candidate combinations related to different reflective surfaces. In the subsequent process of identifying abnormal targets, it can identify abnormal targets caused by different reflective surfaces, without missing any abnormal targets, and the identified abnormal targets are more accurate.

[0107] S104. Select the target combination from the multiple candidate combinations corresponding to each of the multiple detection images.

[0108] In this target combination, the two targets are symmetrical about the reflective surface in the detection image. One target in the target combination is a real target, and the other target is an abnormal target detected by radar based on the reflection of the reflective surface.

[0109] In some embodiments, step S104 can be specifically implemented as follows: the electronic device determines multiple feature parameters corresponding to each target in the detection image based on multiple frames of detection images within a preset time period; then, based on the multiple feature parameters of each target, the electronic device selects a reference combination from multiple candidate combinations, wherein the two targets in the reference combination are symmetrical about the reflective surface in the detection image; for each reference combination, if the ratio of the number of frames containing the reference combination in the multiple frames of detection images to the total number of frames of detection images within the preset time period is greater than a first preset ratio, then the electronic device determines the reference combination as the target combination.

[0110] The reference combination is used to characterize the two targets in the reference combination in the detection image, where one target is suspected to be an abnormal target formed by the other target and the reflective surface.

[0111] For example, if 20 frames of detected images are detected within a preset time period, and candidate combination 1 is determined as a reference combination in 18 of the detected images, and the ratio of the number of frames in which candidate combination 1 is determined as a reference combination to the total number of frames within the preset time period is greater than the first preset ratio of 80%, then candidate combination 1 can be determined as the target combination.

[0112] In this way, by setting the first preset ratio, when the ratio of the number of frames containing the reference combination in the multi-frame detection images to the total number of frames detected within the preset time is greater than the first preset ratio, it indicates that the reference combination is highly credible as the target combination within the preset time. In other words, the electronic device takes into account the movement of the target in the reference combination, and the determined target combination is more credible, and can more accurately determine abnormal targets in the target combination in the subsequent process.

[0113] In some embodiments, the multiple feature parameters include at least region, power, distance, and velocity. Region represents the region corresponding to the target in the detection image; power represents the maximum value of multiple power peaks corresponding to the target in the detection image; distance represents the distance between the target and the radar; and velocity represents the target's velocity. The electronic device selects a reference combination from multiple candidate combinations based on the multiple feature parameters of each target. Specifically, for any candidate combination, the electronic device determines a symmetrical target about the reflector of a first target in the candidate combination. The first target is any target in the candidate combination. If the region corresponding to the symmetrical target in the detection image is similar to that of the second target in the candidate combination... If the area of ​​the overlapping region corresponding to the first target in the detection image is greater than or equal to the area of ​​the second target's region in a ratio greater than or equal to a second preset ratio, the electronic device determines the reflection point between the first and second targets. The second target is a target in the candidate combination other than the first target. When the reflection point is located within the reflector's region, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, the electronic device determines a first proportional relationship between the power of the first target and the distance to the first target, and a second proportional relationship between the power of the second target and the distance to the second target. If the first proportional relationship is different from the second proportional relationship, the electronic device determines that the candidate combination is a reference combination. Furthermore, if the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the second target's region in a ratio less than the second preset ratio, the electronic device determines that the candidate combination is not a reference combination; when the reflection point is not located within the reflector's region, the electronic device determines that the candidate combination is not a reference combination; if the difference between the velocity of the first target and the velocity of the second target is greater than a preset velocity difference, the electronic device determines that the candidate combination is not a reference combination; if the first proportional relationship is the same as the second proportional relationship, the electronic device determines that the candidate combination is not a reference combination.

[0114] The region of the reflector can be understood as the point cloud region corresponding to the reflector.

[0115] An example, such as Figure 7 As shown, a reflector X exists in the detected image. Reflector X has a reflecting surface x. Target A and target B are located on opposite sides of reflecting surface x, and are two targets in a candidate combination. The target A is symmetrical about reflecting surface x. ′ If target A ′ If the area of ​​the region overlapping with target B is 80% of the area of ​​target B, which is greater than the second preset ratio of 70%, then targets A and B can be preliminarily considered to be symmetrical, and the reflection points of targets A and B can be further determined. As before... Figure 7 As shown, the reflection points x of target A and target B are... ′If the velocity of target A is 10 m / s and the velocity of target B is 11 m / s, and the velocity difference between target A and target B is equal to the preset velocity difference of 1 m / s, and the power of target A is inversely proportional to the distance, while the power of target B is directly proportional to the distance, then the electronic device can determine that the candidate combination containing target A and target B is the reference combination, and can proceed to the next step of determining the target combination.

[0116] In another example, a reflector X is detected in the image. Reflector X has a reflecting surface x. Target A and target B are located on opposite sides of reflecting surface x, and are two targets in a candidate combination. Target A is then identified as the symmetrical target A about reflecting surface x. ′ .like Figure 8 As shown in (a), target A ′ If the area of ​​the region overlapping with target B is 10% of the area of ​​target B, which is less than the second preset ratio of 70%, then the electronic device can determine... Figure 8 The candidate combination of target A and target B shown in (a) is not the reference combination. Figure 8 As shown in (b), target A ′ The area of ​​the region overlapping with target B is 80% of the area of ​​target B, which is greater than the second preset ratio of 70%. However, the reflection points x of targets A and B are different. ′ If the object does not fall within the area of ​​reflector X, the electronic device can determine... Figure 8 The candidate combination of target A and target B shown in (b) is not the reference combination. If target A ′ The area of ​​the region overlapping with target B, and the ratio between the area of ​​target A and the area of ​​target B, is greater than a second preset ratio, and the reflection points x of target A and target B are... ′ If the target falls into the area of ​​reflector X, the velocities of target A and target B are then assessed. If the velocity difference between target A and target B is 2 m / s, which is greater than the preset velocity difference of 1 m / s, the electronic equipment can determine that the candidate combination of target A and target B is not the reference combination. If target A... ′ The area of ​​the region overlapping with target B, and the ratio between the area of ​​target A and the area of ​​target B is greater than a second preset ratio, and the reflection points x of target A and target B. ′ If the target falls into the area of ​​reflector X, and the difference between the speeds of target A and target B is less than or equal to the preset speed difference, then it is determined whether the ratio between the distance and power of target A is inversely proportional, and whether the ratio between the distance and power of target B is the same. If both are inversely proportional or both are directly proportional, then the electronic device can determine that the candidate combination composed of target A and target B is not the reference combination.

[0117] Since anomalous targets are formed by the secondary reflection of radar waves reflected from a real target through a reflecting surface, the anomalous target and the real target should be symmetrical about the reflecting surface, and their velocities should be similar. The propagation path of the radar echo signal from one side to the anomalous target is a secondary reflection from the real target through the reflecting surface. Therefore, the propagation time of the anomalous target's echo signal is longer than that of the real target's echo signal. In other words, the distance at which the radar detects the anomalous target should be greater than the distance at which the radar detects the real target. Under normal circumstances, the more reflections and the farther the distance, the weaker the radar wave energy. If a combination contains one real target and one anomalous target, the distance between the real target and the radar should be closer than the distance between the anomalous target and the radar, and the energy of the anomalous target should be weaker than that of the real target. If the distance between target A and the radar in a reference combination is less than the distance between target B and the radar, and the energy of target A is weaker than that of target B, then target A and target B are not a combination of a real target and an anomalous target. Therefore, by constructing a symmetrical target about the reflector of one target in the candidate combination, determining the degree of overlap between the symmetrical target and the other target in the candidate combination, and by determining whether the reflection points of the two targets in the candidate combination fall within the reflector's region, it is possible to accurately determine whether the two targets in the candidate combination are symmetrical. Next, the velocity difference between the two targets in the candidate combination is assessed, and finally, the relationship between the power and distance of each target is determined. This allows for the identification of any suspected anomalous targets in the candidate combination, thus determining whether the candidate combination is a reference combination. In subsequent processes, this also enables a more accurate identification of whether the reference combination is a target combination, filtering out anomalous targets within the target combination.

[0118] In some embodiments, the electronic device determines the error between the point cloud coordinates of the symmetrical target and the point cloud coordinates of the second target. If the ratio between the number of points in the point cloud of the symmetrical target whose point cloud coordinate error with that of the second target is within a preset error range and the total number of points in the point cloud of the second target is greater than or equal to a second preset ratio, then the electronic device determines that the ratio between the area of ​​the overlapping portion of the region corresponding to the symmetrical target in the detection image and the area of ​​the region corresponding to the second target in the detection image in the candidate combination and the area of ​​the region of the second target is greater than or equal to the second preset ratio.

[0119] For example, if there are 10 points in a symmetrical target, for any one of these 10 points, it is determined whether there is a point in the point cloud of the second target whose coordinate error is within a preset error range. If there is, it is determined that this point meets the condition. If there are 9 points in the second target, and 8 out of the 10 points in the symmetrical target meet the condition, then the ratio between these 8 points and the total number of points in the second target is approximately 90%, which is greater than the second preset ratio of 70%. Then the electronic device determines that the area of ​​the overlapping part between the region corresponding to the symmetrical target in the detection image and the region corresponding to the second target in the detection image in the candidate combination is greater than or equal to the area of ​​the region of the second target.

[0120] In this way, by determining the proportion of points in the point cloud of the symmetrical target whose coordinates are within the preset error range from those in the point cloud of the second target to the total number of points in the point cloud of the second target, the degree of overlap between the symmetrical target and the second target can be determined relatively quickly. That is, it can be determined whether the first target and the second target are symmetrical, which is convenient for identifying abnormal targets in the future.

[0121] S105. Identify abnormal targets in the target combination and delete the content information of abnormal targets in the detected image.

[0122] The content information of abnormal targets can be understood as the point cloud data corresponding to the abnormal targets and the feature parameters corresponding to the abnormal targets.

[0123] In some embodiments, the electronic device may delete the content information of abnormal targets in the detected image by deleting points in the point cloud of the abnormal targets in the detected image and deleting the feature parameters corresponding to the abnormal targets presented in the detected image, such as the region, distance, speed, power, etc. of the abnormal targets.

[0124] In some embodiments, the electronic device can determine the abnormal target in the target combination by using multiple feature parameters corresponding to each target in each frame of the detected image.

[0125] For example, the electronic device first judges the average power of the two targets in the target combination. If the average power cannot reflect the difference between the two targets in the target combination, it continues to judge the two targets by combining the average distance. If the average distance also cannot reflect the difference between the two targets in the target combination, it then judges the two targets by combining the length of the target's movement curve within a preset time period, thus identifying the abnormal target in the target combination.

[0126] Specifically, the electronic device determines the abnormal target in the target combination by: determining the average power of the third target and the average power of the fourth target in the target combination. The average power is used to characterize the average power of the target in multiple frames of detected images. The third target is one target in the target combination, and the fourth target is any target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target meet a first preset condition, the electronic device determines the fourth target as an abnormal target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is greater than a first preset difference. If the average power of the third target and the average power of the fourth target do not meet the first preset condition, the electronic device determines the fourth target as an abnormal target in the target combination. The average distance between the third target and the average distance between the fourth target are used to characterize the average distance of the target in the multi-frame detection images. If the average distance between the third target and the average distance between the fourth target meet the second preset condition, the electronic device determines that the fourth target is an abnormal target in the target combination. The second preset condition is that the average distance between the third target is greater than the average distance between the fourth target, and the difference between the average distance between the third target and the average distance between the fourth target is greater than the second preset difference. If the average distance between the third target and the average distance between the fourth target do not meet the second preset condition, the electronic device determines the movement curve of the third target and the movement curve of the fourth target based on the multi-frame detection images. If the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the electronic device determines that the fourth target is an abnormal target in the target combination.

[0127] For example, the target combination includes target A and target B. The average power of target A is a1, and the average power of target B is b1. If a1 > b1, and a1 - b1 is greater than a first preset difference, then target B can be determined to be an abnormal target. If a1 = b1, then the average distance of target A and the average distance of target B are determined. If the average distance between the radar and target A within a preset time period is 10m, the average distance between the radar and target B within a preset time period is 8m, and the second preset difference is 1m, then target B can be determined to be an abnormal target. If the average distance between the radar and target A within a preset time period is 8m, the average distance between the radar and target B within a preset time period is 10m, and the second preset difference is 1m, then target A can be determined to be an abnormal target. If the average distance between the radar and target A within a preset time period is equal to the average distance between the radar and target B within a preset time period, then the electronic device further determines the length of the movement curve of target A within the preset time period, and the length of the movement curve of target B within the preset time period. If the length of the movement curve of target A is greater than the length of the movement curve of target B, the electronic device determines that target B is an abnormal target; if the length of the movement curve of target A is less than the length of the movement curve of target B, the electronic device determines that target A is an abnormal target.

[0128] In this way, based on the difference between abnormal and real targets, the average power of the two targets in the target combination is first determined. The average power of the abnormal target should be less than that of the real target. If the average power cannot reflect the difference between the two targets in the target combination, the average distance is used to further determine the difference between the two targets. The distance between the abnormal target and the radar should be greater than that between the real target and the radar. If the average distance also cannot reflect the difference between the two targets in the target combination, the length of the target's movement curve within a preset time period is used to determine the difference between the two targets. Since the real target is closer to the radar, it will always be detected by the radar first within the preset time period. Therefore, the number of detection frames containing the real target within the preset time period should be greater than the number of detection frames containing the abnormal target. Based on the detection images within the preset time period, it can be determined that the length of the movement curve of the real target is greater than the length of the movement curve of the abnormal target. Through these judgment processes, abnormal targets in the target combination can be accurately identified and filtered out, reducing false detections by the radar, ensuring the accuracy of radar detection, and achieving better application results in subsequent applications of radar detection images.

[0129] In some embodiments, if the average power of the third target and the average power of the fourth target in the target combination do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target do not meet the second preset condition, and the length of the movement curve of the third target is less than or equal to the length of the movement curve of the fourth target, it indicates that the third target and the fourth target are not abnormal targets formed based on the reflective surface from all angles, and the electronic device determines that there are no abnormal targets in the target combination.

[0130] Figure 5 The technical solution described above offers at least the following advantages: By using multiple frames of radar-detected images, reflectors and their reflecting surfaces are identified from among multiple targets in the images. The detected targets are then categorized into two types: a first type located on one side of the reflecting surface and a second type located on the other side. Multiple candidate combinations are generated, each containing one first-type target and one second-type target. From these candidate combinations, a target combination containing one anomalous target and one real target is selected, allowing for the removal of the anomalous target's information. The reflecting surface is determined by combining multiple frames of radar-detected images within a preset time period, taking into account its movement. Based on this, anomalous targets are accurately identified, enabling precise filtering. Based on the working principle of radar, anomalous targets formed by the secondary reflection of target echoes from the reflecting surface must be located on opposite sides of the corresponding real target. Therefore, classifying and combining multiple targets detected by radar based on the reflecting surface, selecting candidate combinations to identify target combinations containing both anomalous and real targets, and then identifying the anomalous target from these combinations is a clear and uncomplicated process, easily reusable in existing equipment. Furthermore, this process only requires radar and does not require additional devices for detecting reflective surfaces, thus saving the cost of devices for filtering out abnormal targets.

[0131] The technical solution provided in this application is described below from the perspective of the overall process:

[0132] (1) Radar detection, to obtain point cloud data and corresponding detection images;

[0133] (2) The electronic device acquires the point cloud data and detection images detected by the radar, clusters the point cloud data, and combines the detection images to obtain multiple targets and the feature parameters of the targets. It records the targets and the feature parameters of the targets represented by each frame of the detection image.

[0134] (3) Based on the feature parameters of multiple targets, the electronic device marks targets that occupy more than a preset number of image blocks in the detected image, have a power greater than a preset power, and have at least one boundary length in the contour that is greater than a preset length as reference targets; targets that are marked as reference targets in multiple frames of detected images within a preset time period, and whose ratio to the total number of frames of detected images within the preset time period is greater than or equal to a third preset ratio, are identified as reflectors; when the ratio of the number of frames of a target marked as reference targets in multiple frames of detected images within a preset time period to the total number of frames of detected images within the preset time period is less than a fourth preset ratio, the electronic device deletes the reflector identifier previously added to the target;

[0135] (4) Fit the reflector and use the fitting result as the reflector surface F; divide the target in the detected image into target K on the left side of the reflector surface and target G on the right side of the reflector surface, generate multiple target pairs based on K and G, and each target pair contains a target in K and a target in G. Traverse each target pair and find the target pair with abnormal targets.

[0136] (5) For any target pair containing target A and target B, target A is mirror-symmetrically transformed according to the reflecting surface F to obtain mirror-symmetric target S. The ratio of the number of point clouds of mirror-symmetric target S and target B in the target pair with coordinate errors within a preset error range to the number of point clouds of target B is calculated. If the ratio is less than a second preset ratio, it can be determined that there is no abnormal target in the target pair, and the next target pair is judged; if the ratio is greater than or equal to the second preset ratio, the electronic device determines the position of the mirror reflection point between target A and target B. If the mirror reflection point is not within the point cloud outline of the reflector, the electronic device determines that there is no abnormal target in the target pair, and the next target pair is judged.

[0137] (6) If the mirror reflection point is within the point cloud outline of the reflector, and the velocity error between target A and target B is greater than the preset velocity difference, the electronic device determines that there is no abnormal target in the target pair and proceeds to the next target pair judgment; if the mirror reflection point is within the point cloud outline of the reflector, and the velocity error between target A and target B is less than or equal to the preset velocity difference, the electronic device determines the proportional relationship between the distance and power of target A and the proportional relationship between the distance and power of target B. If the proportional relationship corresponding to target A is opposite to the proportional relationship corresponding to target B, it is considered that there is an abnormal target formed by the reflection surface F determined in the above steps in target A and target B, and this target pair is determined as a suspected "real target - abnormal target" pair about the reflection surface F; if the proportional relationship corresponding to target A is the same as the proportional relationship corresponding to target B, the electronic device determines that there is no abnormal target in the target pair and proceeds to the next target pair judgment.

[0138] (7) In the multi-frame detection images within a preset time period, the proportion of target pairs marked as suspected “real target - abnormal target” pairs about the reflective surface F is statistically analyzed. If the proportion is greater than the first preset proportion, then target A and target B are determined to be symmetrical about the reflective surface F and are “real target - abnormal target” pairs.

[0139] (8) Compare the average power of target A and target B in the "real target - abnormal target" pair. If the average power of target A is greater than the average power of target B, and the difference in average power between target A and target B is greater than a first preset difference, then the electronic device determines target B as an abnormal target (if the average power of target B is greater than the average power of target A, and the difference in average power between target B and target A is greater than a first preset difference, then the electronic device determines target A as an abnormal target); if the average power of target A and target B are equal, then the electronic device compares the average distance between target A and target B. If the average distance of target A is greater than the average distance of target B, and the difference in average distance between target A and target B is greater than a second preset difference, then the electronic device determines target B as an abnormal target (if the average power of target B is greater than the average power of target A ... B is greater than a second preset difference). If the distance to target A is greater than the average distance to target A, and the difference in average distance between target B and target A is greater than a second preset difference, then the electronic device determines target A as an abnormal target. If the average distances to target A and target B are equal, then the electronic device measures the lengths of the movement curves of target A and target B. If the length of the movement curve of target A is greater than the length of the movement curve of target B, then the electronic device determines target B as an abnormal target. If the lengths of the movement curves of target A and target B are equal, then the electronic device does not consider the target pair consisting of target A and target B to be a "real target - abnormal target" pair, and the electronic device determines that the "real target - abnormal target" judgment in the above steps is incorrect.

[0140] (9) The electronic device deletes the content information of the abnormal target, and then displays the detection image after the content information of the abnormal target has been deleted.

[0141] The foregoing primarily describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the aforementioned functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0142] like Figure 9 As shown in the figure, this application embodiment also provides an abnormal target filtering device for the abnormal target filtering method shown in the above method embodiment. The abnormal target filtering device 400 includes: a processing module 401.

[0143] The processing module 401 is used to: acquire multiple frames of detection images within a preset time period using radar, and determine multiple targets included in the multiple frames of detection images; identify reflectors from the multiple targets, and determine the reflecting surface of the reflectors; classify the multiple targets in each frame of detection images into first-class targets and second-class targets based on the reflecting surface, and generate multiple candidate combinations corresponding to each frame of detection images based on the first-class targets and the second-class targets, wherein a candidate combination includes one first-class target and one second-class target in the same frame of detection images, the first-class target being the target located on one side of the reflecting surface in the detection image, and the second-class target being the target located on the other side of the reflecting surface in the detection image; filter out target combinations from the multiple candidate combinations corresponding to each of the multiple frames of detection images, wherein the two targets in the target combination are symmetrical about the reflecting surface in the detection image, one target in the target combination is a real target, and the other target is an abnormal target detected by radar based on the reflection of the reflecting surface; identify abnormal targets in the target combinations, and delete the content information of abnormal targets in the detection images.

[0144] In one possible implementation, the processing module 401 is specifically used to: determine multiple feature parameters corresponding to each target in each frame of the detection image based on multiple frames of detection images; for each frame of detection image, select a reference combination from multiple candidate combinations corresponding to the detection image based on the multiple feature parameters corresponding to each target in the detection image, wherein the two targets in the reference combination are symmetrical about the reflective surface in the detection image; for each reference combination, if the number of frames in the multiple frames of detection images that contain the reference combination is greater than the ratio of the total number of frames of detection images within a preset time period to a first preset ratio, determine the reference combination as the target combination.

[0145] In another possible implementation, the multiple feature parameters include at least region, power, distance, and velocity. Region represents the region corresponding to the target in the detection image; power represents the maximum value of multiple power peaks corresponding to the target in the detection image; distance represents the distance between the target and the radar; and velocity represents the target's velocity. The processing module 401 is specifically used to: for each candidate combination corresponding to the detection image, determine the symmetrical target of the first target in the candidate combination with respect to the reflector surface, where the first target is any target in the candidate combination; if the area of ​​the overlapping portion between the region corresponding to the symmetrical target in the detection image and the region corresponding to the second target in the detection image in the candidate combination is greater than or equal to the area of ​​the symmetrical target's region, determine the reflection point between the first target and the second target, where the second target is a target in the candidate combination other than the first target; when the reflection point is located within the reflector's region, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, determine a first proportional relationship between the power of the first target and the distance of the first target, and a second proportional relationship between the power of the second target and the distance of the second target; if the first proportional relationship differs from the second proportional relationship, determine that the candidate combination is a reference combination.

[0146] In another possible implementation, the processing module 401 is further configured to: determine that the candidate combination is not a reference combination if the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the region of the symmetrical target; determine that the candidate combination is not a reference combination when the reflection point is not located in the region of the reflector; determine that the candidate combination is not a reference combination if the difference between the velocity of the first target and the velocity of the second target is greater than a preset velocity difference; and determine that the candidate combination is not a reference combination if the first ratio relationship is the same as the second ratio relationship.

[0147] In another possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These feature parameters include at least a region and a power. The region characterizes the region corresponding to the target in the detection image, and the power characterizes the maximum value of multiple power peaks corresponding to the target in the detection image. Specifically, the processing module 401 is used to: determine the average power of the third target and the average power of the fourth target in the target combination. The average power characterizes the average power of the target in multiple frames of the detection image. The third target is one target in the target combination, and the fourth target is any target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target satisfy a first preset condition, the fourth target is determined to be an abnormal target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is large. If the average power of the third target and the average power of the fourth target do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target are determined. The average distance is used to characterize the average value of the distances corresponding to the targets in the multi-frame detection images. If the average distance of the third target and the average distance of the fourth target meet the second preset condition, the fourth target is determined to be an abnormal target in the target combination. The second preset condition is that the average distance of the third target is greater than the average distance of the fourth target, and the difference between the average distance of the third target and the average distance of the fourth target is greater than the second preset difference. If the average distance of the third target and the average distance of the fourth target do not meet the second preset condition, the movement curve of the third target and the movement curve of the fourth target are determined based on the multi-frame detection images. If the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the fourth target is determined to be an abnormal target in the target combination.

[0148] In another possible implementation, each target corresponds to multiple feature parameters in each frame of the detection image. These multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. The processing module 401 is specifically used to: for each frame of the detection image, divide the detection image to obtain multiple non-overlapping image blocks; if the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one side of the region corresponding to the fifth target is greater than a preset length, mark the fifth target as a reference target; if the number of frames in which the fifth target is marked as a reference target in multiple frames of the detection image is greater than or equal to a third preset ratio to the total number of frames of the detection image within a preset time period, determine the fifth target as a reflector; and determine the surface where the side with a length greater than a preset length in the region corresponding to the reflector is determined as the reflecting surface of the reflector.

[0149] It should be noted that, Figure 9The module division shown is illustrative and represents only one logical functional division; in actual implementation, other division methods are possible. For example, two or more functions can be integrated into a single processing module. These integrated modules can be implemented either in hardware or as software functional modules.

[0150] Another embodiment of this application provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the abnormal target filtering method shown in the above embodiments.

[0151] In actual implementation, the processing module 401 can be implemented by the processor of the computer device calling computer program code in memory. The specific execution process can be found in the description of the abnormal target filtering method section above, and will not be repeated here.

[0152] Another embodiment of this application provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the steps of the abnormal target filtering method shown in the above embodiments.

[0153] In another embodiment of this application, a computer program product is also provided, which includes computer instructions that, when executed by a processor, implement the steps of the abnormal target filtering method shown in the above embodiments.

[0154] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), etc.

[0155] The above description is merely a specific embodiment of this application. Any variations or substitutions conceived by those skilled in the art based on the specific embodiments provided in this application should be covered within the protection scope of this application.

Claims

1. An anomaly target filtering method, characterized by, include: Multiple frames of detection images are acquired by radar within a preset time period, and multiple targets included in the multiple frames of detection images are determined. Identify a reflector from the plurality of targets, and determine the reflecting surface of the reflector; Based on the reflective surface, multiple targets in each frame of the detection image are divided into a first type of target and a second type of target. Based on the first type of target and the second type of target, multiple candidate combinations are generated for each frame of the detection image. Each candidate combination includes one first type of target and one second type of target in the same frame of the detection image. The first type of target is the target located on one side of the reflective surface in the detection image, and the second type of target is the target located on the other side of the reflective surface in the detection image. Target combinations are selected from the plurality of candidate combinations corresponding to the multi-frame detection images. The two targets in the target combination are symmetrical about the reflective surface in the detection image. One target in the target combination is a real target, and the other target is an abnormal target detected by the radar based on the reflection of the reflective surface. Identify abnormal targets in the target combination and delete the content information of the abnormal targets in the detected image.

2. The method of claim 1, wherein, The step of selecting the target combination from the plurality of candidate combinations corresponding to each of the multiple detected images includes: Based on the multi-frame detection images, determine multiple feature parameters corresponding to each target in each frame of the detection image; For each frame of the detected image, a reference combination is selected from multiple candidate combinations corresponding to each target in the detected image based on multiple feature parameters corresponding to each target in the detected image. The two targets in the reference combination are symmetrical about the reflective surface in the detected image. For each reference combination, if the number of frames containing the reference combination in the multi-frame detection images is greater than the ratio of the total number of frames in the detection images within the preset time period to a first preset ratio, then the reference combination is determined to be the target combination.

3. The method of claim 2, wherein, The plurality of feature parameters include at least region, power, distance, and velocity. The region is used to characterize the region corresponding to the target in the detection image. The power is used to characterize the maximum value of a plurality of power peaks corresponding to the target in the detection image. The distance is used to characterize the distance between the target and the radar. The velocity is used to characterize the velocity of the target. The step of selecting a reference combination from multiple candidate combinations corresponding to each target in the detected image based on multiple feature parameters includes: For each candidate combination corresponding to the detected image, a symmetrical target of the first target in the candidate combination with respect to the reflecting surface is determined, wherein the first target is any one of the targets in the candidate combination; If the area of ​​the overlapping portion of the region corresponding to the symmetrical target in the detection image and the region corresponding to the second target in the candidate combination in the detection image is greater than or equal to the area of ​​the region of the second target, the reflection point between the first target and the second target is determined, and the second target is a target in the candidate combination other than the first target; When the reflection point is located within the region of the reflector, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, a first proportional relationship between the power of the first target and the distance to the first target, and a second proportional relationship between the power of the second target and the distance to the second target are determined. If the first proportional relationship is different from the second proportional relationship, the candidate combination is determined to be a reference combination.

4. The method of claim 3, wherein, The method further includes: If the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the region of the second target by a ratio less than the second preset ratio, it is determined that the candidate combination is not the reference combination. When the reflection point is not located within the area of ​​the reflector, it is determined that the candidate combination is not the reference combination; If the difference between the speed of the first target and the speed of the second target is greater than the preset speed difference, it is determined that the candidate combination is not the reference combination; If the first proportional relationship is the same as the second proportional relationship, it is determined that the candidate combination is not the reference combination.

5. The method according to claim 1, characterized in that, Each target corresponds to multiple feature parameters in each frame of the detection image. The multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. The process of determining the abnormal targets in the target combination includes: The average power of the third target and the average power of the fourth target in the target combination are determined. The average power is used to characterize the average power of the target in the multi-frame detection images. The third target is a target in the target combination, and the fourth target is a target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target satisfy the first preset condition, the fourth target is determined to be an abnormal target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is greater than the first preset difference. If the average power of the third target and the average power of the fourth target do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target are determined. The average distance is used to characterize the average value of the distances corresponding to the targets in the multi-frame detection images. If the average distance of the third target and the average distance of the fourth target meet the second preset condition, the fourth target is determined to be an abnormal target in the target combination. The second preset condition is that the average distance of the third target is greater than the average distance of the fourth target, and the difference between the average distance of the third target and the average distance of the fourth target is greater than the second preset difference. If the average distance between the third target and the average distance between the fourth target do not meet the second preset condition, the movement curves of the third target and the fourth target are determined based on the multi-frame detection images. If the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the fourth target is determined to be an abnormal target in the target combination.

6. The method according to any one of claims 1-5, characterized in that, Each target corresponds to multiple feature parameters in each frame of the detection image. The multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. The step of determining a reflector from the plurality of targets and determining the reflecting surface of the reflector includes: For each frame of the detected image, the detected image is divided into multiple non-overlapping image blocks; If the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one side of the region corresponding to the fifth target is greater than a preset length, the fifth target is marked as a reference target; If the number of frames in the multi-frame detection images in which the fifth target is marked as the reference target is greater than or equal to the total number of frames in the detection images within the preset time period, then the fifth target is determined to be the reflector. The surface containing the edge with a length greater than the preset length in the region corresponding to the reflector is defined as the reflective surface of the reflector.

7. An anomaly target filtering device characterized by comprising: Includes a processing module, the processing module being used for: Multiple frames of detection images are acquired by radar within a preset time period, and multiple targets included in the multiple frames of detection images are determined. Identify a reflector from the plurality of targets, and determine the reflecting surface of the reflector; Based on the reflective surface, multiple targets in each frame of the detection image are divided into a first type of target and a second type of target. Based on the first type of target and the second type of target, multiple candidate combinations are generated for each frame of the detection image. Each candidate combination includes one first type of target and one second type of target in the same frame of the detection image. The first type of target is the target located on one side of the reflective surface in the detection image, and the second type of target is the target located on the other side of the reflective surface in the detection image. Target combinations are selected from the plurality of candidate combinations corresponding to the multi-frame detection images. The two targets in the target combination are symmetrical about the reflective surface in the detection image. One target in the target combination is a real target, and the other target is an abnormal target detected by the radar based on the reflection of the reflective surface. Identify abnormal targets in the target combination and delete the content information of the abnormal targets in the detected image.

8. The apparatus according to claim 7, characterized in that, The processing module is specifically used for: determining multiple feature parameters corresponding to each target in each frame of the detection image based on the multi-frame detection images; for each frame of the detection image, selecting a reference combination from multiple candidate combinations corresponding to the detection image based on the multiple feature parameters corresponding to each target in the detection image, wherein the two targets in the reference combination are symmetrical about the reflective surface in the detection image; for each reference combination, if the number of frames in the multi-frame detection images containing the reference combination is greater than a first preset ratio to the total number of frames in the detection images within the preset time period, the reference combination is determined to be the target combination; The plurality of feature parameters include at least region, power, distance, and velocity. The region is used to characterize the region corresponding to the target in the detection image. The power is used to characterize the maximum value of a plurality of power peaks corresponding to the target in the detection image. The distance is used to characterize the distance between the target and the radar. The velocity is used to characterize the velocity of the target. The processing module is specifically used to: for each candidate combination corresponding to the detection image, determine a symmetrical target of the first target in the candidate combination with respect to the reflector surface. The first target is any target in the candidate combination. If the area of ​​the overlapping portion of the region corresponding to the symmetrical target in the detection image and the region corresponding to the second target in the candidate combination in the detection image is greater than or equal to the area of ​​the region of the second target, a reflection point between the first target and the second target is determined, wherein the second target is a target in the candidate combination other than the first target; when the reflection point is located within the region of the reflector, and the difference between the velocity of the first target and the velocity of the second target is less than or equal to a preset velocity difference, a first proportional relationship between the power of the first target and the distance to the first target, and a second proportional relationship between the power of the second target and the distance to the second target are determined; if the first proportional relationship is different from the second proportional relationship, the candidate combination is determined to be a reference combination; The processing module is further configured to: determine that the candidate combination is not a reference combination if the area of ​​the region corresponding to the symmetrical target in the detection image is less than the area of ​​the region of the second target by a ratio smaller than the second preset ratio; determine that the candidate combination is not a reference combination when the reflection point is not located within the region of the reflector; determine that the candidate combination is not a reference combination when the difference between the velocity of the first target and the velocity of the second target is greater than the preset velocity difference; and determine that the candidate combination is not a reference combination when the first ratio is the same as the second ratio. Each target corresponds to multiple feature parameters in each frame of the detected image. These multiple feature parameters include at least a region and a power. The region characterizes the region corresponding to the target in the detected image, and the power characterizes the maximum value of multiple power peaks corresponding to the target in the detected image. Specifically, the processing module is used to: determine the average power of the third target and the average power of the fourth target in the target combination. The average power characterizes the average power of the target in the multiple frames of the detected image. The third target is one target in the target combination, and the fourth target is any target in the target combination other than the third target. If the average power of the third target and the average power of the fourth target satisfy a first preset condition, the fourth target is determined to be an abnormal target in the target combination. The first preset condition is that the average power of the third target is greater than the average power of the fourth target, and the difference between the average power of the third target and the average power of the fourth target is greater than a first preset difference. If the average power of the third target and the average power of the fourth target do not meet the first preset condition, the average distance of the third target and the average distance of the fourth target are determined. The average distance is used to characterize the average value of the distances corresponding to the targets in the multi-frame detection images. If the average distance of the third target and the average distance of the fourth target meet the second preset condition, the fourth target is determined to be an abnormal target in the target combination. The second preset condition is that the average distance of the third target is greater than the average distance of the fourth target, and the difference between the average distance of the third target and the average distance of the fourth target is greater than the second preset difference. If the average distance between the third target and the average distance between the fourth target do not meet the second preset condition, the movement curves of the third target and the fourth target are determined based on the multi-frame detection images. If the length of the movement curve of the third target is greater than the length of the movement curve of the fourth target, the fourth target is determined to be an abnormal target in the target combination; Each target corresponds to multiple feature parameters in each frame of the detection image. The multiple feature parameters include at least a region and a power. The region is used to characterize the region corresponding to the target in the detection image, and the power is used to characterize the maximum value of multiple power peaks corresponding to the target in the detection image. The processing module is specifically used to: for each frame of the detection image, divide the detection image to obtain multiple non-overlapping image blocks; if the number of image blocks occupied by the region of the fifth target in the detection image is greater than a preset number, the power of the fifth target is greater than a preset power, and the length of at least one side of the region corresponding to the fifth target is greater than a preset length, mark the fifth target as a reference target. If the number of frames in the multi-frame detection images in which the fifth target is marked as the reference target is greater than or equal to the total number of frames in the detection images within the preset time period, then the fifth target is determined to be the reflector. The surface containing the edge with a length greater than the preset length in the region corresponding to the reflector is defined as the reflective surface of the reflector.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-6.

10. A computer readable storage medium having stored thereon computer instructions, wherein, When executed by a processor, the computer instructions implement the steps of the method according to any one of claims 1-6.

11. A computer program product comprising computer instructions, characterized in that, When executed by a processor, the computer instructions implement the steps of the method according to any one of claims 1-6.

Images