Target recognition method, device, electronic device and storage medium

Abstract

The present invention relates to the field of radar target recognition technology, and provides a target recognition method, device, electronic device and storage medium. The method is applied to an electronic device, which is connected to a millimeter wave radar and pre-stores a historical data table, which includes the identification and recognition results of historically identified targets; by first parsing the data reported by the millimeter wave radar, the original data of the target to be identified and the identification in the original data are obtained; if the identification of the target to be identified exists in the historical data table, the first recognition result of the target to be identified is obtained according to the historical data table; if the identification of the target to be identified does not exist in the historical data table, the second recognition result of the target to be identified is obtained according to the original data of the target to be identified and a preset recognition model. Identification based on historically identified targets and recognition models can reduce the error rate of target recognition, thereby improving the accuracy and efficiency of recognition.

Description

Technical Field

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

[0002] Millimeter wave radar has the characteristics of stable detection performance, long range, good environmental applicability, and strong ability to penetrate fog, smoke, and dust. Therefore, millimeter wave radar is widely used in traffic target detection scenarios. It can collect and track the traffic status of road vehicles and guide traffic flow. However, in complex intersection scenarios, there are cases where radar tracking targets are lost and radar recognition targets are wrong. Summary of the invention

[0003] In view of this, an object of the present invention is to provide a target recognition method, device, electronic device and storage medium.

[0004] In order to achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0005] In a first aspect, the present invention provides a target recognition method, which is applied to an electronic device, wherein the electronic device is connected to a millimeter wave radar for communication, and the electronic device pre-stores a historical data table, wherein the historical data table includes an identification and a recognition result of a target recognized historically; the method comprises:

[0006] Parsing the data reported by the millimeter wave radar to obtain the original data of the target to be identified; the original data includes an identifier;

[0007] If the identifier of the target to be identified exists in the historical data table, obtaining a first identification result of the target to be identified according to the historical data table;

[0008] If the identifier of the target to be identified does not exist in the historical data table, a second identification result of the target to be identified is obtained according to the original data of the target to be identified and a preset identification model.

[0009] In an optional implementation, the historical data table includes a first data table and a second data table, the first data table includes identifiers of historically identified false alarm targets, and the second data table includes a mapping relationship between post-splitting identifiers and pre-splitting identifiers of historically identified split targets;

[0010] The step of obtaining a first recognition result of the target to be recognized according to the historical data table comprises:

[0011] If the identifier of the target to be identified exists in the first data table, the first identification result is a false alarm target;

[0012] If the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified exists in the second data table, the first identification result is a split target;

[0013] If the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified does not exist in the second data table, the first identification result is a normal target.

[0014] In an optional implementation manner, the electronic device further pre-stores a target data table, and the target data table caches the original data of N historical targets;

[0015] The step of obtaining a second recognition result of the target to be identified based on the original data of the target to be identified and a preset recognition model comprises:

[0016] For N historical targets, the i-th eigenvalue set is calculated according to the original data of the i-th historical target and the original data of the target to be identified, so as to obtain N eigenvalue sets; N is a positive integer, and i is a positive integer not greater than N;

[0017] According to the N feature value sets and the recognition model, a second recognition result of the target to be identified is obtained; the second recognition result is a false alarm target, a split target, or a normal target.

[0018] In an optional implementation, the raw data includes reporting time, coordinates, speed and acceleration;

[0019] The step of calculating the i-th eigenvalue set according to the original data of the i-th historical target and the original data of the target to be identified comprises:

[0020] Obtaining an i-th time difference according to the reporting time of the i-th historical target and the reporting time of the target to be identified;

[0021] Obtaining an i-th distance difference according to the coordinates of the i-th historical target and the coordinates of the target to be identified;

[0022] According to the i-th distance difference and the i-th time difference, the i-th distance change rate is obtained;

[0023] Obtaining an i-th speed difference according to the speed of the i-th historical target and the speed of the target to be identified;

[0024] According to the i-th speed difference and the i-th time difference, the i-th speed change rate is obtained;

[0025] Obtaining an i-th acceleration difference according to the acceleration of the i-th historical target and the acceleration of the target to be identified;

[0026] According to the ith acceleration difference and the ith time difference, the acceleration change rate is obtained to obtain the ith eigenvalue; wherein the eigenvalue set includes time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate.

[0027] In an optional implementation manner, the electronic device further pre-stores a target data table, the target data table caches original data of N historical targets; the second recognition result is a false alarm target, or a split target, or a normal target; the method further includes:

[0028] According to the second recognition result, updating the historical data table and / or the target data table;

[0029] The step of updating the historical data table and / or the target data table according to the second recognition result includes:

[0030] If the second recognition result is a false alarm target, updating the identifier of the target to be identified to the first data table;

[0031] If the second recognition result is a split target, the identifier of the target to be identified is used as a post-splitting identifier, the pre-splitting identifier is obtained, and the post-splitting identifier and the corresponding pre-splitting identifier are updated to the second data table; and after the identifier of the target to be identified is modified to the corresponding pre-splitting identifier, the original data of the target to be identified is updated to the target data table;

[0032] If the second recognition result is a normal target, the original data of the target to be recognized is updated to the target data table.

[0033] In an optional embodiment, the recognition model is obtained in the following manner:

[0034] Acquire raw data and labels of multiple radar targets, wherein the labels are used to indicate whether the radar targets are false alarm targets, split targets, or normal targets;

[0035] Obtaining, according to the original data of the multiple radar targets, a feature set of each radar target to be processed among the multiple radar targets;

[0036] According to the feature sets and labels of all radar targets to be processed, the preset basic model is trained to obtain the recognition model.

[0037] In an optional implementation, the raw data includes a reporting time; and the step of obtaining a feature set of each to-be-processed radar target among the multiple radar targets according to the raw data of the multiple radar targets includes:

[0038] For each of the radar targets to be processed, original data of N historical radar targets before the reporting time of the radar target to be processed are obtained;

[0039] Obtaining N feature value sets according to the original data of the radar target to be processed and the original data of the N historical radar targets; the feature set includes the N feature value sets;

[0040] A feature set of each radar target to be processed is obtained.

[0041] In a second aspect, the present invention provides a target recognition device, which is applied to an electronic device, wherein the electronic device is communicatively connected to a millimeter wave radar, and the electronic device pre-stores a historical data table, wherein the historical data table includes an identification and a recognition result of a target recognized historically; the device includes:

[0042] A parsing module, used to parse the data reported by the millimeter wave radar to obtain the original data of the target to be identified; the original data includes an identifier;

[0043] A first identification module, configured to obtain a first identification result of the target to be identified according to the historical data table if the identifier of the target to be identified exists in the historical data table;

[0044] The second recognition module is used to obtain a second recognition result of the target to be recognized based on the original data of the target to be recognized and a preset recognition model if the identifier of the target to be recognized does not exist in the historical data table.

[0045] In a third aspect, the present invention provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in any one of the aforementioned embodiments is implemented.

[0046] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any one of the aforementioned embodiments.

[0047] The target recognition method, device, electronic device and storage medium provided by the embodiment of the present invention are applied to an electronic device, which is connected to the millimeter wave radar and pre-stores a historical data table, which includes the identification and recognition results of the targets identified in the past; the original data of the target to be identified and the identification in the original data are obtained by first parsing the data reported by the millimeter wave radar; if the identification of the target to be identified exists in the historical data table, the first recognition result of the target to be identified is obtained according to the historical data table; if the identification of the target to be identified does not exist in the historical data table, the second recognition result of the target to be identified is obtained according to the original data of the target to be identified and a preset recognition model. Identification based on historically identified targets and recognition models can reduce the error rate of target recognition, thereby improving the accuracy and efficiency of recognition.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 A block diagram of an electronic device provided by an embodiment of the present invention is shown;

[0051] Figure 2 A schematic diagram of a process flow of a target recognition method provided by an embodiment of the present invention is shown;

[0052] Figure 3 Another schematic diagram of a target recognition method provided by an embodiment of the present invention is shown;

[0053] Figure 4 Another schematic diagram of a target recognition method provided by an embodiment of the present invention is shown;

[0054] Figure 5 Another schematic diagram of a target recognition method provided by an embodiment of the present invention is shown;

[0055] Figure 6 Another schematic diagram of a target recognition method provided by an embodiment of the present invention is shown;

[0056] Figure 7 Another schematic diagram of a target recognition method provided by an embodiment of the present invention is shown;

[0057] Figure 8 A functional module diagram of a target recognition device provided by an embodiment of the present invention is shown.

[0058] Icon: 110 - bus; 120 - processor; 130 - memory; 170 - communication interface; 300 - target recognition device; 310 - parsing module; 330 - first recognition module; 350 - second recognition module; 370 - training module. DETAILED DESCRIPTION

[0059] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0060] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0061] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0062] Millimeter wave radar refers to a radar with a wavelength of 1mm-10mm and a frequency of 10GHZ-200GHZ. It has the characteristics of stable detection performance, long range, good environmental applicability, small size, light weight and high spatial resolution. Millimeter wave radar has a strong ability to penetrate fog, smoke and dust, and can work all day and all weather. Compared with laser radar, its cost is low. Therefore, it is widely used in traffic target detection scenarios. By collecting and tracking the traffic status of road vehicles, it provides basic data for signal control state and potential perception, and assists traffic police in traffic diversion and timing design for traffic flow.

[0063] However, due to the basic principle of target positioning and speed measurement, namely the Doppler effect, it is easy to lose the target when tracking vehicles in tangential motion and vehicles moving slowly at a low speed. For example, in a complex intersection scene, because the vehicle is controlled by the traffic light and is waiting in line, the vehicle moves slowly or stops, which can easily cause the radar to lose the tracking target, causing the radar to identify the same vehicle as a different target. From the perspective of the visualized target trajectory, the trajectory of the vehicle is divided into multiple non-overlapping sub-trajectories, which is a target with split trajectory. At the same time, at the intersection of the road, due to the influence of pedestrians, non-motor vehicles or other non-stable moving reflective surfaces, the radar will receive a reflected echo, causing the radar to identify a target at the reflected position, which is a false alarm target. Since the accuracy of radar target recognition is low in complex traffic scenes, the embodiment of the present invention provides a target recognition method to solve the above technical problems.

[0064] Please refer to Figure 1 , is a block diagram of an electronic device provided by an embodiment of the present invention. The electronic device includes a bus 110 , a processor 120 , a memory 130 , and a communication interface 170 .

[0065] The bus 110 may be a circuit that connects the above elements to each other and transfers communications (eg, control messages) between the above elements.

[0066] The processor 120 may receive commands from the other elements described above (eg, the memory 130 , the communication interface 170 , etc.) through the bus 110 , may interpret the received commands, and may perform calculations or data processing according to the interpreted commands.

[0067] The processor 120 may be an integrated circuit chip with signal processing capabilities. The processor 120 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0068] The memory 130 may store commands or data received from the processor 120 or other elements (eg, the communication interface 170 , etc.) or commands or data generated by the processor 120 or other elements.

[0069] The memory 130 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0070] The communication interface 170 may be used to communicate signaling or data with other node devices.

[0071] Understandably, Figure 1 The structure shown is only a schematic diagram of the structure of the electronic device. The electronic device may also include Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown. Figure 1 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0072] The above-mentioned electronic device will be used as an execution subject to execute each step in each method provided in the embodiment of the present invention and achieve corresponding technical effects.

[0073] See also Figure 2 , Figure 2 : is a flow chart of a target recognition method provided by an embodiment of the present invention. It can be understood that the method provided by an embodiment of the present invention can be applied to the scene of traffic target detection for tracking vehicles. The electronic device is connected to the millimeter wave radar for communication, and the millimeter wave radar detects the target on the traffic road and reports the collected target data to the electronic device, and the electronic device recognizes the target based on the data.

[0074] The electronic device pre-stores a historical data table and a preset recognition model. The millimeter-wave radar can report some data within a preset time period, or report data of a set number of targets. The recognition model recognizes the data reported by the millimeter-wave radar to obtain the recognition result of the target. The historical data table is used to record the identification and recognition results of the historically recognized targets.

[0075] Step S202, parsing the data reported by the millimeter wave radar to obtain the original data of the target to be identified;

[0076] The original data includes an identifier; the target to be identified can be understood as the target currently reported by the millimeter-wave radar to the electronic device. In order to distinguish different detected targets, the millimeter-wave radar includes a unique identifier for characterizing the target in the data reported.

[0077] The data reported by the millimeter wave radar and received by the electronic device may be data encoded based on a coding protocol, which may be a string of characters encoded in hexadecimal format.

[0078] Optionally, the data reported by the millimeter wave radar can be parsed according to a coding protocol pre-stored in the electronic device to obtain the original data of the target to be identified, which includes an identifier, that is, the identifier of the target to be identified is obtained.

[0079] For ease of understanding, the following Table 1 is an example of a parsed data field provided by an embodiment of the present invention. As shown in Table 1, the target number is the identifier. It should be noted that the coding protocol of the radar may be different based on different radar manufacturers, and the embodiment of the present invention does not limit the coding protocol of the radar.

[0080] Table 1

[0081]

[0082]

[0083] Step S204, if the identification of the target to be identified exists in the historical data table, obtaining a first identification result of the target to be identified according to the historical data table;

[0084] Optionally, the historical data table records the identification and recognition results of the targets identified historically. The historical data table can be queried to see whether there is a representation of the target to be identified. If so, the recognition result of the target to be identified, i.e., the first recognition result, can be obtained based on the historical data table; if not, step S206 can be executed.

[0085] It can be understood that if the identification of the target to be identified exists in the historical data table, directly obtaining the identification result from the historical data table can quickly and accurately identify the target to be identified, thereby improving the efficiency and accuracy of the identification.

[0086] Step S206: if the identifier of the target to be identified does not exist in the historical data table, a second identification result of the target to be identified is obtained according to the original data of the target to be identified and a preset identification model.

[0087] Among them, the preset recognition model is a model pre-trained based on a large amount of data, which can recognize the target. The method of obtaining the recognition model will be introduced in the subsequent steps.

[0088] Optionally, the historical data table includes a first data table, a second data table and a third data table, wherein the first data table and the second data table are used to record the identification of abnormal targets such as false alarm targets and split targets identified in the history; and the third data table is used to record the identification of normal targets identified in the history.

[0089] Optionally, if the identification of the target to be identified does not exist in the historical data table, it means that the target to be identified is not a target that has been identified historically but a new target. The target to be identified can be identified through the recognition model, that is, the recognition result of the target to be identified, that is, the second recognition result, can be obtained based on the original data of the target to be identified and the recognition model.

[0090] It can be seen that based on the above steps, the electronic device is connected to the millimeter wave radar for communication and a historical data table is pre-stored in the electronic device, the historical data table includes the identification and recognition results of the historically recognized targets; the original data of the target to be recognized and the identification in the original data are obtained by first parsing the data reported by the millimeter wave radar; if the identification of the target to be recognized exists in the historical data table, the first recognition result of the target to be recognized is obtained according to the historical data table; if the identification of the target to be recognized does not exist in the historical data table, the second recognition result of the target to be recognized is obtained according to the original data of the target to be recognized and the preset recognition model. Recognition based on historically recognized targets and recognition models can reduce the error rate of target recognition, thereby improving the accuracy and efficiency of recognition.

[0091] Optionally, if there is an identifier of the target to be identified in the historical data table, it means that the target to be identified is a target that has been processed by historical identification. The historical data table may include a first data table and a second data table, and these two data tables may be used to record abnormal targets identified historically. The first data table may be used to record the identifier of the false alarm target identified historically, and the second data table may be used to record the mapping relationship between the post-splitting identifier and the pre-splitting identifier of the split target identified historically. Furthermore, the embodiment of the present invention provides a possible implementation method for the above-mentioned step S204, wherein step S204 may include the following steps:

[0092] Step S204a, if the identifier of the target to be identified exists in the first data table, the first identification result is a false alarm target; wherein the first data table includes the identifiers of the false alarm targets identified historically.

[0093] The identification of the target to be identified can be queried in the first data table to determine whether there is an identification of the target to be identified. If there is, the first identification result of the target to be identified is a false alarm target, which can be understood as the target detected by the millimeter wave radar is not the target to be tracked. If not, it is determined whether there is an identification of the target to be identified in the second data table.

[0094] For example, the target that the millimeter-wave radar detects is actually a vehicle, but due to the interference of pedestrians, the millimeter-wave radar detects a target, namely, the target to be identified, at the position of the pedestrian. If the identifier of the target to be identified is found in the first data table, the target to be identified is a false alarm target.

[0095] Optionally, if the first recognition result of the target to be identified is a false alarm target, the original data of the target to be identified is discarded, that is, subsequent data processing is not performed on the false alarm target.

[0096] Step S204b, if the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified exists in the second data table, the first identification result is a split target; wherein the second data table includes a mapping relationship between the post-splitting identifier and the pre-splitting identifier of the split target identified historically.

[0097] If the identifier of the target to be identified does not exist in the first data table, the identifier of the target to be identified can be queried in the second data table to determine whether it exists. If it exists, the first identification result of the target to be identified is a split target, which can be understood as the target detected by the millimeter-wave radar is actually a target detected before. If it does not exist, the first identification result of the target to be identified is a normal target.

[0098] For example, the millimeter-wave radar has previously detected target A, i.e., vehicle A. Vehicle A stops at an intersection waiting for a traffic light. When vehicle A starts driving again, the millimeter-wave radar detects it as target B, i.e., the target to be identified.

[0099] If the mapping relationship between the identifier of target B and the identifier of target A is recorded in the second data table, the identifier of target B indicates the identifier after the splitting, and the identifier of target A indicates the identifier before the splitting. The identifier before the splitting can be obtained according to the identifier after the splitting, that is, the identifier of target A can be obtained according to the identifier of target B. That is, if the identifier of the target to be identified, i.e., target B, exists in the second data table, it means that the target to be identified, i.e., target B, is a split target, and it can be determined that target B is actually target A that has been detected before.

[0100] Optionally, if the first identification result of the target to be identified is a split target, the identifier of the target to be identified can be used as the post-splitting identifier, and the pre-splitting identifier corresponding to the post-splitting identifier can be obtained. After the identifier of the target to be identified is modified to the corresponding pre-splitting identifier, the original data of the target to be identified can be applied to subsequent radar tracking target business scenarios.

[0101] Step S204c: if the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified does not exist in the second data table, the first identification result is a normal target.

[0102] If the identifier of the target to be identified does not exist in the first data table and the second data table, and the identifier of the target to be identified exists in the third data table, and the third data table is used to record the identifiers of normal targets identified historically, then the target to be identified is a normal target identified historically.

[0103] Optionally, if the first recognition result of the target to be identified is a normal target, the original data of the target to be identified may be applied to a subsequent business scenario of radar tracking the target.

[0104] It is understandable that the electronic device pre-stores a target data table, which is used to cache data reported by the millimeter wave radar within a preset time period, or data of a set number of targets. The target to be identified is the currently detected target, and in order to facilitate the identification of the next detected target, the target data table needs to be updated.

[0105] If the target to be identified is a split target, the identifier of the target to be identified is modified to the corresponding pre-splitting identifier, and the original data of the target to be identified is updated to the target data table. If the target to be identified is a normal target, the original data of the target to be identified is directly updated to the target data table.

[0106] Optionally, if there is no identification of the target to be identified in the historical data table, it means that the target to be identified is not a target that has been identified historically, and a second identification result of the target to be identified can be obtained based on the original data of the target to be identified and the identification model. Furthermore, the embodiment of the present invention provides a possible implementation method for the above step S206. Step S206 may include the following steps:

[0107] The electronic device may pre-store a target data table, where the target data table is used to cache the latest original data of the N historical targets, where the original data of the N historical targets are arranged in sequence according to the reporting time.

[0108] Step S206-1, for N historical targets, calculate the i-th eigenvalue set according to the original data of the i-th historical target and the original data of the target to be identified, to obtain N eigenvalue sets; wherein N is a positive integer, and i is a positive integer not greater than N;

[0109] Optionally, the original data of N historical targets before the target to be identified that are cached in the target data table can be used to calculate feature values ​​based on the original data of the target to be identified and the original data of each historical target, and a set of N feature values ​​can be obtained. The feature value can be understood as the feature change of the target to be identified relative to the historical target in the motion dimension.

[0110] Step S206-3, obtaining a second recognition result of the target to be recognized based on the N feature value sets and the recognition model; wherein the second recognition result is a false alarm target, a split target, or a normal target.

[0111] Optionally, the obtained N feature value sets can be input into the recognition model to obtain a second recognition result of the target to be recognized. The second recognition result may be a false alarm target, that is, the target to be recognized is a false alarm target; it may also be a split target, that is, the target event to be recognized is a target that has been recognized in the past; the target to be recognized may be a normal target.

[0112] Optionally, in order to further improve the accuracy of recognition, the N feature value sets may be preprocessed, such as data missing value processing, outlier analysis, duplicate data filtering, feature collinearity processing, data standardization and normalization processing, etc.

[0113] For ease of understanding, the present invention provides a flowchart. Figure 3 , the above steps are introduced below with reference to the figure.

[0114] In step S202, the electronic device parses the data reported by the millimeter wave radar to obtain the original data of the target to be identified, and obtains the identification of the target to be identified;

[0115] In step 204-1, it is determined whether there is an identifier of the target to be identified in the first data table. If so, the first identification result of the target to be identified is a false alarm target, that is, step 204a; if not, step S204-3 is executed.

[0116] In step S204-3, it is determined whether there is an identifier of the target to be identified in the target data table. If so, step S204-5 is executed; if not, step S206-1 is executed.

[0117] In step S204-5, it is determined whether there is an identifier of the target to be identified in the second data table. If so, the first identification result of the target to be identified is a split target, i.e., step 204b; if not, the first identification result of the target to be identified is a normal target, i.e., step 204c.

[0118] In step S206-1, based on the original data of N historical targets cached in the target data table, each feature value set is calculated according to the original data of each historical target and the original data of the target to be identified, to obtain N feature value sets.

[0119] In step S206-3, a second recognition result of the target to be recognized is obtained according to the N feature value sets and the recognition model. The second recognition result of the target to be recognized may be a false alarm target, a split target, or a normal target.

[0120] It can be seen that the target recognition method provided by the embodiment of the present invention can recognize the target to be recognized based on the target recognized in history, or can recognize the target to be recognized based on the recognition model. Through the two recognition methods, the recognition result can be obtained quickly, and the efficiency and accuracy of target recognition are improved.

[0121] Optionally, for the above step S206-1, the embodiment of the present invention provides a possible implementation method. Figure 4 , wherein step S206-1 may further include the following steps. It is understandable that, Figure 4 The execution order of steps S206-1-2 to step S206-1-14 shown in the figure is only an example of the embodiment of the present invention, and the execution steps can be designed according to actual applications, and the embodiment of the present invention is not limited thereto.

[0122] In Table 1 given in the above step S202, the original data includes the data sending time, ie, the reporting time, the X coordinate and the Y coordinate, ie, the coordinate, the vehicle speed, ie, the speed, and the acceleration in the moving direction, ie, the acceleration.

[0123] Step S206-1-2, obtaining the i-th time difference according to the reporting time of the i-th historical target and the reporting time of the target to be identified;

[0124] Optionally, the absolute difference between the reporting time of the target to be identified and the reporting time of each historical target may be calculated to obtain each time difference.

[0125] Step S206-1-4, obtaining the i-th distance difference according to the coordinates of the i-th historical target and the coordinates of the target to be identified;

[0126] Optionally, the absolute value of the distance difference between the coordinates of the target to be identified and the coordinates of each historical target may be calculated to obtain each distance difference.

[0127] Step S206-1-6, obtaining the i-th distance change rate according to the i-th distance difference and the i-th time difference;

[0128] Optionally, the ratio of the i-th distance difference to the i-th time difference may be used as the i-th distance change rate to obtain each distance change rate.

[0129] Step S206-1-8, obtaining the i-th speed difference according to the speed of the i-th historical target and the speed of the target to be identified;

[0130] Optionally, the absolute difference between the speed of the target to be identified and the speed of each historical target may be calculated to obtain each speed difference.

[0131] Step S206-1-10, obtaining the i-th speed change rate according to the i-th speed difference and the i-th time difference;

[0132] Optionally, the ratio of the ith speed difference to the ith time difference may be used as the ith speed change rate to obtain each speed change rate.

[0133] Step S206-1-12, obtaining the i-th acceleration difference according to the acceleration of the i-th historical target and the acceleration of the target to be identified;

[0134] Optionally, the absolute difference between the acceleration of the target to be identified and the acceleration of each historical target may be calculated to obtain each speed difference.

[0135] Step S206-1-14, obtain the acceleration change rate according to the i-th acceleration difference and the i-th time difference, and obtain the i-th eigenvalue set; wherein the eigenvalue set includes time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate.

[0136] Optionally, the ratio of the ith acceleration difference to the ith time difference can be used as the ith acceleration change rate to obtain each acceleration change rate. The obtained time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate can be used as a feature value of the target to be identified, and N feature value sets can be obtained.

[0137] Optionally, based on N historical radar targets, the target to be identified may have N+1 feature value sets, wherein the N+1th feature value set may include N discrete statistical parameters of time difference, N discrete statistical parameters of distance difference, N discrete statistical parameters of distance change rate, N discrete statistical parameters of speed difference, N discrete statistical parameters of speed change rate, N discrete statistical parameters of acceleration, and N discrete statistical parameters of acceleration change rate. The discrete statistical parameters may be mean values ​​and / or variances, and / or standard deviations, etc.

[0138] Optionally, in order to facilitate the identification of the next target, after obtaining the second identification result of the target to be identified, the historical data table and / or the target data table may be updated according to the second identification result. Figure 5 After step S206-3, the following steps may also be included:

[0139] Step S208-1, if the second recognition result is a false alarm target, updating the identifier of the target to be recognized to the first data table;

[0140] Optionally, if the second recognition result of the target to be recognized is a false alarm target, the identification of the target to be recognized is updated to the first data table, so as to use the latest first data table to detect whether the identification of the next target is a false alarm, which can improve the recognition speed and accuracy.

[0141] Step S208-3, if the second recognition result is a split target, the identifier of the target to be identified is used as the post-splitting identifier, the pre-splitting identifier is obtained, and the post-splitting identifier and the corresponding pre-splitting identifier are updated to the second data table; and, after the identifier of the target to be identified is modified to the corresponding pre-splitting identifier, the original data of the target to be identified is updated to the target data table;

[0142] Optionally, if the second recognition result of the target to be identified is a split target, the identifier of the target to be identified may be used as a post-splitting identifier, and a corresponding pre-splitting identifier may be determined from historical targets cached in the target data table.

[0143] The method for obtaining the pre-splitting identification can be to obtain the positions of the first N historical targets, namely the X-coordinate and the Y-coordinate, the speed, namely the X-axis speed and the Y-axis speed, the acceleration, namely the X-axis acceleration and the Y-axis acceleration, and the time difference, namely the time difference between the reporting time of the split target and the reporting time of each historical target; then according to the kinematic formula, for example, according to the uniformly accelerated motion formula, the position of the split target is predicted based on the N historical targets to obtain N predicted positions; then, the distance is calculated based on the N predicted positions and the current position of the split target, and the historical target corresponding to the predicted position with the shortest distance is taken as the target before splitting, the identification of the historical target is obtained, the pre-splitting identification is obtained, and a mapping relationship between the post-splitting identification and the pre-splitting identification is established.

[0144] Updating the post-splitting identifier and the corresponding pre-splitting identifier into the second data table establishes a mapping relationship between the two, so as to use the latest second data table to detect whether the identifier of the next target is split, which can improve the speed and accuracy of recognition.

[0145] In addition, the identifier of the target to be identified is modified to the corresponding identifier before splitting, and then the original data of the target to be identified is updated to the target data table, so as to calculate the characteristic value of the next target using the latest target data table.

[0146] Step S208-5: If the second recognition result is a normal target, the original data of the target to be recognized is updated to the target data table.

[0147] Optionally, if the second recognition result of the target to be identified is a normal target, the original data of the target to be identified is updated to the target data table, so as to calculate the characteristic value of the next target using the latest target data table.

[0148] It can be seen that based on the second recognition result obtained by the recognition model, the historical data table and the target data table can be updated according to the second recognition result, so that subsequent targets can be identified through feedback, thereby achieving rapid recognition and improving the accuracy of target recognition.

[0149] In the above-described embodiment, the recognition model provided by the embodiment of the present invention can identify false alarm targets and split targets. In the related art, false alarm targets can be identified based on radar positioning and tracking principles, from the mode of radar echo mixing basic data processing. For example, the side lobe is processed by windowing mode, and the RDM with stronger contrast between the target and the background noise is obtained, and the noise background and the target are separated, and the constant false alarm detection is adopted to adaptively estimate the surrounding noise background of each point, and then the intensity of each point is compared with the surrounding intensity, and finally determine whether the point is the target of tracking. But adopting this method, good effect can not be obtained for objects with strong reflection surfaces, and it is impossible to identify whether it is a split target.

[0150] Alternatively, the target data reported by the radar can be processed to identify the split target. For example, when the target is slow or stopped, the radar cannot collect target information. Kalman filtering and other target position prediction methods can be used to predict the target position after the radar loses the target. The position of the target reacquired by the radar is matched with the previously predicted position. If the two match, they are considered to be the same target. However, in this way, if the radar does not collect data for a long time, there will be a deviation between the predicted position and the actual position. The longer the time, the greater the deviation, and it is impossible to identify whether it is a false alarm target.

[0151] It can be seen that both of these two methods cannot achieve the recognition of both false alarm targets and split targets. Therefore, the embodiment of the present invention provides a recognition model that can recognize false alarm targets and split targets. Figure 6 , the following will combine Figure 6 The method of obtaining the recognition model is introduced.

[0152] Step S212, obtaining raw data and labels of multiple radar targets; wherein the labels are used to indicate whether the radar target is a false alarm target, a split target, or a normal target;

[0153] It is understandable that obtaining the recognition model requires pre-collection of radar target data. In the embodiment of the present invention, data can be collected based on the millimeter wave radar, the shooting device and the terminal device such as a computer, and the three are connected to each other for communication.

[0154] For example, a radar and a camera can be installed at an intersection. During the radar installation process, calibration is required to ensure that the location information of the target reported by the radar is accurate. Then the camera is installed so that the shooting range of the camera coincides with the detection range of the radar, so that the radar target can be marked in subsequent steps.

[0155] After installation, synchronize the millimeter-wave radar, camera, and computer time. Then, during the morning rush hour, such as 8:30 to 9:30, when the traffic volume is large and vehicles are prone to queues and congestion, it is convenient to collect data on false alarm targets, split targets, and normal targets. Collect data through radar and turn on the camera to shoot the traffic road in real time. The computer is used to record the collected radar data.

[0156] After acquiring the radar data, the coding protocol can be used to decode the radar data to obtain the original data of multiple radar targets.

[0157] Then, multiple radar targets are made into visual videos. You can find the recorded road section on the map, obtain the base map of the road section, convert the longitude and latitude coordinates of the target collected by the radar into the pixel position of the base map, mark the target with pixel blocks, and make video frames; set the number of frames per second of the video playback, continuously read the recorded radar data, and make a visual video of the radar target.

[0158] It should be noted that in order to ensure that the duration of the produced visualization video is consistent with the duration of the collected radar data, the time cannot be stretched or compressed during the process of producing the video. An embodiment of the present invention provides a possible implementation method.

[0159] You can set the display frequency to K frames per second. K is generally lower than the frequency at which the radar sends data. K should not be too low. If the number of frames per second is too low, the video will have a noticeable jump feeling. The FPS of the video produced is K, that is, the difference between two frames is The first frame is T0, which is the earliest time reported in the recorded radar data, and the radar target at that time is displayed for a duration of seconds; The time is the second frame, and the radar target displayed can be Time range target, displayed for seconds. At the third frame, the radar target displayed can be Time range target, displayed for seconds, ... The time is the sth frame, and so on, until the recorded radar data processing is completed.

[0160] It is worth noting that when a larger K value is set, the time interval between two frames In seconds, not all moving targets in the radar field of view will be detected and reported by the radar, which means that the target may not appear in a certain frame, but may appear in adjacent frames, and the final reflection in the visual video is that the target flickers. To avoid this situation, when making the sth frame, you can select a radar target with a longer time range to make the video frame, such as For all targets that appear within the time range, deduplicate the targets, that is, only retain the data of the last appearance, and make video frames.

[0161] The radar visualization video and the video footage taken by the camera are played synchronously, and the radar target is matched with the vehicle target in the video footage one by one, and the radar target is marked. It can obtain three types of labels: false alarm target, split target and normal target.

[0162] For example, vehicle target P1 starts to appear in the radar field of view and drives away from the radar field of view, then target P1 is a normal target. Vehicle target P2 suddenly disappears during the movement and soon appears as target P3 at another location, but target P2 and target P3 are the same vehicle in the video recording, then target P3 is a split target. There is no vehicle in the video recording, and it appears as target P4 in the radar visualization video, then target P4 is a false alarm target. Each radar target can be marked to obtain its corresponding label, as shown in Table 2, a data table based on this example.

[0163] Table 2

[0164] Logo Label P1 Normal target P3 Split Target P4 False alarm target

[0165] The labeler can play two videos at the same time, match the radar target in the visualization with the vehicle target in the video recording one by one, and mark the radar target to obtain labels of multiple radar targets. The label can be used to indicate whether the radar target is a false alarm target, a split target, or a normal target.

[0166] Step S214, obtaining a feature set of each radar target to be processed in the multiple radar targets according to the original data of the multiple radar targets;

[0167] Optionally, the raw data of the radar targets can be arranged in chronological order based on the reporting time in the raw data of the radar targets, and the raw data of the radar targets from the earliest reporting time to a preset time length can be pre-selected, or the raw data of a preset number of radar targets from the earliest reporting time can be selected.

[0168] The method of obtaining the feature set in the embodiment of the present invention needs to be calculated based on the original data of the radar target currently being processed and the radar target before it, and these pre-selected radar targets are used as initial radar targets to calculate the feature value set of subsequent radar targets. Some radar targets are selected from multiple radar targets as initial radar targets, and the others are all radar targets to be processed. The radar target to be processed can be understood as a radar target whose identification appears for the first time except the initial radar target.

[0169] Optionally, a feature set of each radar target to be processed may be obtained according to the original data of multiple radar targets.

[0170] Step S216, training the preset basic model according to the feature sets and labels of all radar targets to be processed to obtain a recognition model.

[0171] Optionally, after obtaining the feature set of each radar target to be processed, a preset basic model may be trained according to the feature sets and labels of all radar targets to be processed to obtain a recognition model.

[0172] In order to obtain a better performance of the recognition model, the feature set of the radar target to be processed can be preprocessed in advance. For example, label matching processing, matching the feature set of the radar target to be processed with its label. Duplicate data filtering processing. Feature collinearity processing, deleting feature values ​​with high collinearity. Data standardization and normalization processing.

[0173] Data missing value processing: Due to missing data, some feature values ​​in the feature value sets of some radar targets are missing. In this case, the missing values ​​need to be processed. Depending on the amount of data and the number of missing fields, the optional processing methods include: deleting records, filling records, etc.

[0174] Outlier analysis and processing: analyze each characteristic value separately and check the abnormal data. The abnormal data can be processed as missing values, delete outliers, truncate outliers to specific values, etc.

[0175] Unbalanced processing of category data. Depending on the time periods selected for labeling, the three types of labels, namely false alarm targets, split targets, and normal targets, may be highly unbalanced. For example, when there are fewer vehicles, there are fewer false alarm targets and split targets. During congested periods, there are more records of split targets. Unbalanced labels need to be processed before model training. Common processing methods include undersampling of classes with more records, oversampling of classes with more records, and adjustments when selecting algorithms and designing loss functions.

[0176] All radar targets to be processed after preprocessing are divided into training radar targets and test radar targets according to the preset ratio, such as 7:3 or 8:2. The targets with a larger ratio are used as training radar targets, and the targets with a smaller ratio are used as test radar targets. The original data of the training radar targets are used to train the preset basic model, and the original data of the test radar targets are used to evaluate the trained preset basic model. You can choose three algorithms: xgboost, catboot, and lightgbm for training.

[0177] You can select evaluation indicators of the classification model, such as precision, recall, F1 value, and AUC value, to evaluate the trained model. Based on multiple evaluation indicators, you can adjust and optimize the model in a targeted manner, such as optimizing model parameters, optimizing data preprocessing methods, and changing training algorithms, so that the model performance reaches the best.

[0178] For the above step S214, the embodiment of the present invention provides a possible implementation method. Figure 7 , wherein step S214 may include the following steps:

[0179] Step S214-1, for each radar target to be processed, obtaining the original data of N historical radar targets before the reporting time of the radar target to be processed.

[0180] It is understandable that a queue can be used to cache the initial radar target in the above steps, and a time window T is set to record the time length of the cached raw data, and a queue Q is set to record the latest n raw data of the same radar target. If the radar reports a new raw data, the new raw data is inserted into the head of the corresponding queue, and the historical raw data in the queue is automatically moved to the tail of the queue. If the historical raw data in the queue is full, the raw data at the tail of the queue is automatically discarded. At the same time, when the reporting time of the latest raw data in the queue to the current time exceeds the time window T, all the raw data cached in the queue are automatically discarded.

[0181] The latest original data of N historical radar targets before the reporting time of the radar target to be processed can be obtained through the time pane T and the N queues Q.

[0182] Step S214-3, obtaining N feature value sets according to the original data of the pending radar target and the original data of N historical radar targets; wherein the feature set includes N feature value sets.

[0183] It is understandable that the raw data includes reporting time, coordinates, speed and acceleration.

[0184] Optionally, the time difference can be obtained by the reporting time of each historical radar target and the reporting time of the radar target to be processed. The distance difference is obtained by the coordinates of each historical radar target and the coordinates of the radar target to be processed; the distance change rate is obtained based on the distance difference and the time difference. The speed difference is obtained by the speed of each historical radar target and the speed of the radar target to be processed; the speed change rate is obtained based on the speed difference and the time difference. The acceleration difference is obtained by the acceleration of each historical radar target and the acceleration of the radar target to be processed; the acceleration change rate is obtained based on the acceleration difference and the time difference.

[0185] It should be noted that when calculating eigenvalues ​​based on raw data, we should try to avoid using specific spatial or temporal information such as longitude and latitude and specific time to calculate eigenvalues. We should try to make the constructed features as general as possible so that the model can learn the relationship patterns between targets and improve the generalization ability of the model.

[0186] Based on the raw data of the radar target to be processed and the raw data of a historical radar target, a feature value set can be obtained, which includes time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate. Based on N historical radar targets, N feature value sets can be obtained. These N feature value sets are feature sets.

[0187] Optionally, based on N historical radar targets, the radar target to be determined can also obtain N+1 characteristic values. The N+1th characteristic value can be N discrete statistical parameters of time difference, N discrete statistical parameters of distance difference, N discrete statistical parameters of distance change rate, N discrete statistical parameters of speed difference, N discrete statistical parameters of speed change rate, N discrete statistical parameters of acceleration and N discrete statistical parameters of acceleration change rate. The discrete statistical parameters can be mean value and / or variance, and / or standard deviation, etc.

[0188] It is understandable that the number of feature value sets constructed for a radar target in the training model process should be consistent with the number of feature value sets constructed for a radar target in the use model process. For example, in the training mode process, ten feature value sets are constructed with a radar target and its previous ten radar targets. In the use model process, ten feature value sets are also required to be constructed with a radar target and its previous ten radar targets to ensure that the recognition model can accurately identify the target.

[0189] Step S214-5, obtaining a feature set of each radar target to be processed;

[0190] Optionally, for each radar target to be processed, steps S214 - 1 to S214 - 3 are executed to obtain a feature set of each radar target to be processed.

[0191] In order to execute the corresponding steps in the above embodiments and various possible methods, a method for implementing a target recognition device is given below. Figure 8 , Figure 8 This is a functional module diagram of a target recognition device 300 provided in an embodiment of the present invention. It should be noted that the basic principle and technical effects of the target recognition device 300 provided in this embodiment are the same as those of the above embodiment. For the sake of brief description, for parts not mentioned in this embodiment, reference can be made to the corresponding contents in the above embodiment. The target recognition device 300 includes:

[0192] The parsing module 310 is used to parse the data reported by the millimeter wave radar to obtain the original data of the target to be identified; the original data includes an identifier;

[0193] A first identification module 330, configured to obtain a first identification result of the target to be identified according to the historical data table if the identifier of the target to be identified exists in the historical data table;

[0194] The second recognition module 350 is used to obtain a second recognition result of the target to be recognized based on the original data of the target to be recognized and a preset recognition model if the identifier of the target to be recognized does not exist in the historical data table.

[0195] Optionally, the first identification module 330 is specifically used for: if the identification of the target to be identified exists in the first data table, the first identification result is a false alarm target; if the identification of the target to be identified does not exist in the first data table and the identification of the target to be identified exists in the second data table, the first identification result is a split target; if the identification of the target to be identified does not exist in the first data table and the identification of the target to be identified does not exist in the second data table, the first identification result is a normal target.

[0196] Optionally, the second identification module 350 is specifically used to: for N historical targets, calculate the i-th eigenvalue set based on the original data of the i-th historical target and the original data of the target to be identified, to obtain N eigenvalue sets; N is a positive integer, and i is a positive integer not greater than N; based on the N eigenvalue sets and the identification model, obtain a second identification result of the target to be identified; the second identification result is a false alarm target, a split target, or a normal target.

[0197] Optionally, the second identification module 350 is specifically used to: obtain the ith time difference according to the reporting time of the ith historical target and the reporting time of the target to be identified; obtain the ith distance difference according to the coordinates of the ith historical target and the coordinates of the target to be identified; obtain the ith distance change rate according to the ith distance difference and the ith time difference; obtain the ith speed difference according to the speed of the ith historical target and the speed of the target to be identified; obtain the ith speed change rate according to the ith speed difference and the ith time difference; obtain the ith acceleration difference according to the acceleration of the ith historical target and the acceleration of the target to be identified; obtain the acceleration change rate according to the ith acceleration difference and the ith time difference, and obtain the ith eigenvalue set; wherein the eigenvalue set includes time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate.

[0198] Optionally, the second identification module 350 is further used to update the historical data table and / or the target data table according to the second identification result.

[0199] Optionally, the second identification module 350 is also used to: if the second identification result is a false alarm target, update the identifier of the target to be identified to the first data table; if the second identification result is a split target, use the identifier of the target to be identified as the post-splitting identifier, obtain the pre-splitting identifier, and update the post-splitting identifier and the corresponding pre-splitting identifier to the second data table; and, after modifying the identifier of the target to be identified to the corresponding pre-splitting identifier, update the original data of the target to be identified to the target data table; if the second identification result is a normal target, update the original data of the target to be identified to the target data table.

[0200] The target recognition device 300 also includes a training module 370, which is used to: obtain original data and labels of multiple radar targets, where the labels are used to indicate whether the radar target is a false alarm target, a split target, or a normal target; obtain a feature set of each radar target to be processed in the multiple radar targets based on the original data of the multiple radar targets; and train a preset basic model based on the feature sets and labels of all radar targets to be processed to obtain a recognition model.

[0201] The training module 370 is specifically used for: for each radar target to be processed, obtaining the original data of N historical radar targets before the reporting time of the radar target to be processed; obtaining N feature value sets based on the original data of the radar target to be processed and the original data of the N historical radar targets; the feature set includes N feature value sets; and obtaining the feature set of each radar target to be processed.

[0202] An embodiment of the present invention further provides an electronic device, including a processor 120 and a memory 130. The memory 130 stores a computer program. When the processor executes the computer program, the target recognition method disclosed in the above embodiment is implemented.

[0203] The embodiment of the present invention further provides a storage medium on which a computer program is stored. When the computer program is executed by the processor 120, the target recognition method disclosed in the embodiment of the present invention is implemented.

[0204] In summary, the target recognition method, device, electronic device and storage medium provided by the embodiments of the present invention are applied to an electronic device, which is connected to the millimeter wave radar and pre-stores a historical data table, which includes the identification and recognition results of the targets identified historically; the original data of the target to be identified and the identification in the original data are obtained by first parsing the data reported by the millimeter wave radar; if the identification of the target to be identified exists in the historical data table, the first recognition result of the target to be identified is obtained according to the historical data table; if the identification of the target to be identified does not exist in the historical data table, the second recognition result of the target to be identified is obtained according to the original data of the target to be identified and a preset recognition model. Identification based on historically identified targets and recognition models can reduce the error rate of target recognition, thereby improving the accuracy and efficiency of recognition.

[0205] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0206] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0207] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0208] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A target recognition method, characterized in that: Applied to an electronic device, the electronic device is connected to a millimeter wave radar for communication, the electronic device pre-stores a historical data table, the historical data table includes the identification and recognition results of historically recognized targets, the electronic device also pre-stores a target data table, the target data table caches the original data of N historical targets; the method includes: Parsing the data reported by the millimeter wave radar to obtain the original data of the target to be identified; the original data includes identification, reporting time, coordinates, speed and acceleration; If the identifier of the target to be identified exists in the historical data table, obtaining a first identification result of the target to be identified according to the historical data table; If the identifier of the target to be identified does not exist in the historical data table, obtaining a second identification result of the target to be identified according to the original data of the target to be identified and a preset identification model; The step of obtaining a second recognition result of the target to be identified based on the original data of the target to be identified and a preset recognition model comprises: For N historical targets, according to the reporting time of the i-th historical target and the reporting time of the target to be identified, obtain the i-th time difference; Obtaining an i-th distance difference according to the coordinates of the i-th historical target and the coordinates of the target to be identified; According to the i-th distance difference and the i-th time difference, the i-th distance change rate is obtained; Obtaining an i-th speed difference according to the speed of the i-th historical target and the speed of the target to be identified; According to the i-th speed difference and the i-th time difference, the i-th speed change rate is obtained; Obtaining an i-th acceleration difference according to the acceleration of the i-th historical target and the acceleration of the target to be identified; According to the i-th acceleration difference and the i-th time difference, the acceleration change rate is obtained, the i-th eigenvalue set is obtained, and N eigenvalue sets are obtained; the eigenvalue set includes time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate; the eigenvalue set represents the characteristic change of the target to be identified relative to the historical target in the motion dimension; N is a positive integer, and i is a positive integer not greater than N; The N feature value sets are input into the recognition model to obtain a second recognition result of the target to be identified; the second recognition result is a false alarm target, a split target, or a normal target; the false alarm target indicates a target that is not actually tracked by the millimeter-wave radar, and the split target indicates the same target that is identified as a different target by the millimeter-wave radar.

2. The method according to claim 1, characterized in that The historical data table includes a first data table and a second data table, the first data table includes identifiers of historically identified false alarm targets, and the second data table includes a mapping relationship between post-splitting identifiers and pre-splitting identifiers of historically identified split targets; The step of obtaining a first recognition result of the target to be recognized according to the historical data table comprises: If the identifier of the target to be identified exists in the first data table, the first identification result is a false alarm target; If the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified exists in the second data table, the first identification result is a split target; If the identifier of the target to be identified does not exist in the first data table and the identifier of the target to be identified does not exist in the second data table, the first identification result is a normal target.

3. The method according to claim 2, characterized in that The electronic device also pre-stores a target data table, and the target data table caches the original data of N historical targets; The second recognition result is a false alarm target, a split target, or a normal target; and the method further includes: According to the second recognition result, updating the historical data table and / or the target data table; The step of updating the historical data table and / or the target data table according to the second recognition result includes: If the second recognition result is a false alarm target, updating the identifier of the target to be identified to the first data table; If the second recognition result is a split target, the identifier of the target to be identified is used as a post-splitting identifier, the pre-splitting identifier is obtained, and the post-splitting identifier and the corresponding pre-splitting identifier are updated to the second data table; and after the identifier of the target to be identified is modified to the corresponding pre-splitting identifier, the original data of the target to be identified is updated to the target data table; If the second recognition result is a normal target, the original data of the target to be recognized is updated to the target data table.

4. The method according to claim 1, characterized in that: The recognition model is obtained in the following manner: Acquire raw data and labels of multiple radar targets, wherein the labels are used to indicate whether the radar targets are false alarm targets, split targets, or normal targets; Obtaining, according to the original data of the multiple radar targets, a feature set of each radar target to be processed among the multiple radar targets; According to the feature sets and labels of all radar targets to be processed, the preset basic model is trained to obtain the recognition model.

5. The method according to claim 4, characterized in that The original data includes a reporting time; the step of obtaining a feature set of each radar target to be processed among the multiple radar targets according to the original data of the multiple radar targets includes: For each of the radar targets to be processed, obtaining original data of N historical radar targets before the reporting time of the radar target to be processed; Obtaining N feature value sets according to the original data of the radar target to be processed and the original data of the N historical radar targets; the feature set includes the N feature value sets; A feature set of each radar target to be processed is obtained.

6. A target recognition device, characterized in that: Applied to an electronic device, the electronic device is connected to a millimeter wave radar for communication, the electronic device pre-stores a historical data table, the historical data table includes the identification and recognition results of historically recognized targets, the electronic device also pre-stores a target data table, the target data table caches the original data of N historical targets; the device includes: A parsing module, used to parse the data reported by the millimeter wave radar to obtain the original data of the target to be identified; the original data includes an identifier; A first identification module, configured to obtain a first identification result of the target to be identified according to the historical data table if the identifier of the target to be identified exists in the historical data table; A second recognition module, configured to obtain a second recognition result of the target to be recognized based on the original data of the target to be recognized and a preset recognition model if the identifier of the target to be recognized does not exist in the historical data table; The second recognition module is also used for: for N historical targets, obtaining the ith time difference according to the reporting time of the ith historical target and the reporting time of the target to be identified; obtaining the ith distance difference according to the coordinates of the ith historical target and the coordinates of the target to be identified; obtaining the ith distance change rate according to the ith distance difference and the ith time difference; obtaining the ith speed difference according to the speed of the ith historical target and the speed of the target to be identified; obtaining the ith speed change rate according to the ith speed difference and the ith time difference; obtaining the ith acceleration difference according to the acceleration of the ith historical target and the acceleration of the target to be identified; obtaining the acceleration according to the ith acceleration difference and the ith time difference. Speed ​​change rate, obtain the i-th eigenvalue set, and obtain N eigenvalue sets; the eigenvalue sets include time difference, distance difference, distance change rate, speed difference, speed change rate, acceleration difference and acceleration change rate; the eigenvalue set represents the characteristic change of the target to be identified relative to the historical target in the motion dimension; N is a positive integer, and i is a positive integer not greater than N; the N eigenvalue sets are input into the recognition model to obtain a second recognition result of the target to be identified; the second recognition result is a false alarm target, a split target, or a normal target; the false alarm target represents a target that is not actually tracked by the millimeter-wave radar, and the split target represents the same target recognized as different targets by the millimeter-wave radar.

7. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images