Target classification method, device, equipment and storage medium

Abstract

The present application relates to the field of target analysis, and in particular to a target classification method, device, equipment and storage medium. The method comprises: determining a tracking target; obtaining M continuous historical classification labels obtained by the latest classification of the tracking target, wherein M is an integer greater than 1; determining, based on the M historical classification labels, that labels frequently switch when classifying the tracking target; obtaining a target classification label, and performing switching delay processing on the tracking target through the target classification label. The present application is used to solve the problem that when classifying a target, classification labels frequently switch, resulting in an uneven classification process.

Description

Technical Field

[0001] The present application relates to the field of target analysis, and in particular to a target classification method, device, equipment and storage medium. Background Art

[0002] With the development of society, automatic classification of targets in the environment has become an important issue. For example, when counting the number of people in a space, how to identify the human targets in the space. At present, the technical basis of the existing target classification monitoring system includes infrared technology, laser technology and image processing technology.

[0003] However, when the existing technology classifies targets, there is a problem of frequent switching of classification labels due to objective reasons such as the target recognition angle. For example, when identifying people in a space through image processing technology, due to the problem of image acquisition accuracy, the label for classifying the target frequently switches between human and non-human, and the classification process is not smooth, resulting in the inability to obtain accurate classification labels, which in turn causes a reduction in classification accuracy. Summary of the invention

[0004] The present application provides a target classification method, apparatus, device and storage medium, which are used to solve the problem of frequent switching of classification labels when classifying targets, resulting in an uneven classification process.

[0005] In a first aspect, an embodiment of the present application provides a target classification method, including: determining a tracking target; obtaining M continuous historical classification labels obtained from the latest classification of the tracking target, where M is an integer greater than 1; determining frequent label switching when classifying the tracking target based on the M historical classification labels; obtaining a target classification label, and performing switching delay processing on the tracking target through the target classification label.

[0006] Optionally, the acquiring of the M consecutive historical classification labels obtained by the latest classification of the tracking target includes: the M consecutive historical classification labels obtained by the latest classification of the tracking target within a switching time threshold; the determining, based on the M historical classification labels, that labels are frequently switched when classifying the tracking target includes: acquiring a number of label switching times based on the M historical classification labels; when the number of label switching times is greater than a switching number threshold, it is determined that labels are frequently switched when classifying the tracking target.

[0007] Optionally, obtaining the target classification label includes: obtaining N consecutive historical classification labels obtained by the latest classification of the tracking target, wherein N is an integer greater than 1; calculating the proportion of each historical classification label among the N historical classification labels; comparing the proportion of each historical classification label with a preset proportion threshold corresponding to each historical classification label; and using the historical classification label whose proportion is greater than the preset proportion threshold as the target classification label.

[0008] Optionally, determining the tracking target includes: obtaining a millimeter-wave radar echo signal of the detection target; parsing the echo signal to obtain a current distance of the target from the millimeter-wave radar; when the current distance is less than a first distance threshold, determining the target as the tracking target.

[0009] Optionally, the obtaining of the latest continuous M historical classification labels for the tracking target includes: tracking the trajectory of the tracking target through the millimeter-wave radar; obtaining the echo signal of the millimeter-wave radar during the trajectory tracking process; parsing the echo signal of the millimeter-wave radar during the trajectory tracking process through a machine learning classification algorithm, and continuously classifying the tracking target; obtaining the latest continuous M historical classification labels for the tracking target.

[0010] Optionally, performing the switching delay processing on the tracked target by using the target classification label includes: keeping the latest classification result of the tracked target as the target classification label unchanged until the keeping time reaches a preset delay time length.

[0011] Optionally, when the holding time reaches a preset delay time, the method further includes: reclassifying the tracking target until it is determined again based on the M historical classification labels that labels are frequently switched when classifying the tracking target.

[0012] In a second aspect, an embodiment of the present application provides a target classification device, comprising: a first determination module, used to determine a tracking target; an acquisition module, used to obtain M consecutive historical classification labels obtained by the latest classification of the tracking target, wherein M is an integer greater than 1; a second determination module, used to determine frequent label switching when classifying the tracking target based on the M historical classification labels; a delay processing module, used to obtain a target classification label, and perform switching delay processing on the tracking target through the target classification label.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to implement the target classification method described in the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the target classification method described in the first aspect.

[0015] Compared with the prior art, the above technical solution provided by the embodiment of the present application has the following advantages: After determining the tracking target, the method provided by the embodiment of the present application determines when the tracking target frequently switches labels during classification based on the M continuous historical classification labels obtained by the latest classification of the tracking target, and then obtains the target classification label, and performs switching delay processing on the tracking target through the target classification label. Through this method, when tracking and classifying the tracking target, the continuous and frequent switching of the classification label is avoided. When frequent switching occurs, delay processing is immediately performed to obtain the target classification label after delay processing, so as to achieve smooth optimization of the classification process and thereby improve the classification accuracy. Avoid poor usage effects caused by frequent label switching and improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 Schematic diagram of the process steps for implementing the target classification method provided in the embodiment of the present application Figure 1 ;

[0019] Figure 2 Schematic diagram of the process steps for determining whether labels are frequently switched when classifying a tracking target provided in an embodiment of the present application Figure 1 ;

[0020] Figure 3 The following is a schematic diagram of the specific implementation method steps of the target classification method provided in the embodiment of the present application. Figure 2 ;

[0021] Figure 4The following is a schematic diagram of the specific implementation method steps of the target classification method provided in the embodiment of the present application. Figure 3 ;

[0022] Figure 5 Schematic diagram of the steps for determining whether the tracking target has frequent label switching provided in the embodiment of the present application Figure 2 ;

[0023] Figure 6 This is a schematic diagram of the structure connection of the target classification device provided in the embodiment of the present application;

[0024] Figure 7 This is a schematic diagram of the structural connection of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0026] The target classification method provided in the embodiment of the present application is used to classify the target. The method is implemented in a device for performing the target classification method, and the device can be a smart phone, a tablet computer, a smart camera, or any other device capable of classifying the target. The scope of protection of the present application is not limited to the specific implementation form of the device.

[0027] In one embodiment, if Figure 1 As shown, the process steps for implementing the target classification method are as follows:

[0028] Step 101: determine the tracking target.

[0029] In this embodiment, the classification of the target is a continuous tracking process, and the tracking target refers to the target that needs to be classified. That is, after the tracking target is determined, the tracking target will be continuously classified to obtain at least one classification label.

[0030] In this embodiment, the tracking target can be determined by infrared technology, laser technology, image processing technology, millimeter wave radar technology, and any other technology that can implement this solution. The protection scope of this application is not limited to the specific technology for determining the tracking target.

[0031] In one embodiment, the process of determining the tracking target uses millimeter wave radar technology. Based on the millimeter wave radar technology, the specific process of determining the tracking target is as follows: obtaining the millimeter wave radar echo signal of the detection target; parsing the echo signal to obtain the current distance of the target from the millimeter wave radar; when the current distance is less than the first distance threshold, determining the target as the tracking target.

[0032] In this embodiment, millimeter waves refer to electromagnetic waves with a wavelength of 1 to 10 mm, which are located in the wavelength range where microwaves and far-infrared waves overlap, and thus have the characteristics of both spectra. After a millimeter wave transmitting device transmits millimeter waves into a space, the targets in the space will reflect echoes, and the millimeter wave receiving device can receive the echoes. By analyzing the echo signals, information such as the location and distance of the space target can be obtained.

[0033] During the use of millimeter waves, compared with centimeter wave seekers, millimeter wave seekers have the characteristics of small size, light weight and high spatial resolution. Compared with optical seekers such as infrared, laser, and television, millimeter wave seekers have strong ability to penetrate fog, smoke, and dust, and have the characteristics of all-weather (except heavy rainy days) and all-day. In addition, the anti-interference and anti-stealth capabilities of millimeter wave seekers are also better than other microwave seekers.

[0034] In this embodiment, the millimeter wave transmitting device and the receiving device may be an integrated millimeter wave chip, which may be installed in a device for classifying tracked targets, so as to implement the target classification method of the present application through millimeter wave radar technology. It should be noted that the millimeter wave transmitting device and the receiving device may also be implemented in other forms or installed in other locations according to actual conditions and needs, as long as the target classification method of the present application can be implemented.

[0035] In this embodiment, the first distance threshold is a preset distance threshold. After obtaining the current distance of the target from the parsed echo signal, if the current distance is less than the first distance threshold, the target is determined to be a tracking target. The specific value of the first distance threshold is set according to actual conditions and needs.

[0036] The first distance threshold is used to clarify the conditions for determining the target to be tracked, avoiding the situation where all targets are tracked and classified unconditionally, resulting in classification confusion. By adjusting the specific value of the first distance threshold, it is possible to ensure that a higher-precision echo signal is obtained, thereby improving the accuracy of the entire target classification.

[0037] Step 102: Obtain M consecutive historical classification labels obtained by the latest classification of the tracking target, where M is an integer greater than 1.

[0038] In this embodiment, the classification label refers to the result obtained by classifying the tracking target, representing the type of the tracking target. The specific classification label can be set in advance according to needs and actual conditions. For example, the classification label can include human and non-human, and the classification label can also include human, cat and dog. The protection scope of this application is not limited to the specific number and specific content of the classification label.

[0039] In this embodiment, since the classification of the tracking target is a continuous process, the historical classification label refers to the classification label obtained by classifying the tracking target before the current moment. The M historical classification labels are the latest and continuous historical classification labels obtained before the current moment, which can ensure the timeliness of determining whether the classification label is frequently switched, and avoid the situation where the historical classification label used is long away from the current moment, the time period label of the historical classification label is frequently switched, but the label of the current moment is no longer frequently switched, thereby ensuring the timeliness and accuracy of target classification.

[0040] In one embodiment, when it is determined that the tracking target adopts millimeter wave radar technology, at least one historical classification label of the tracking target is obtained based on the millimeter wave radar technology, and the specific process is as follows: the trajectory of the tracking target is tracked by the millimeter wave radar; the echo signal of the millimeter wave radar in the trajectory tracking process is obtained; the echo signal of the millimeter wave radar in the trajectory tracking process is analyzed by a machine learning classification algorithm, and the tracking target is continuously classified; and M continuous historical classification labels obtained by the latest classification of the tracking target are obtained.

[0041] Specifically, based on millimeter wave radar technology, after collecting the echo signal of the tracking target, the echo signal is converted into digital form; the digital signal after the analog-to-digital conversion is processed by digital signal processing (DSP) to obtain point cloud information. Each frame of the point cloud signal point cloud information contains multiple point information, where the point information is a vector composed of distance, azimuth, Doppler velocity information and signal-to-noise ratio. When the point cloud information is 3D point cloud information, it also contains height information. Then, the trajectory tracking of the tracking target is realized according to the adaptive extended Kalman filter algorithm.

[0042] Based on the point cloud information of the tracking target, feature extraction is performed on the point cloud information through preset classification features, such as preset human body features, and the extracted features are input into a preset machine learning classification algorithm to achieve classification of the tracking target.

[0043] In this embodiment, the point cloud information of the target can be obtained through the echo signal of the millimeter-wave radar. Based on the advantages of the millimeter-wave radar, the trajectory tracking is more convenient to implement, and the feature information obtained is more rapid and accurate, thereby improving the accuracy of target classification.

[0044] Step 103: determining, based on the M historical classification labels, whether labels are frequently switched when classifying the tracking target.

[0045] In this embodiment, label switching refers to the classification result changing from one label to another. For example, if the classification label obtained in the first classification is human, and the classification label obtained in the second classification is non-human, this process is a label switching.

[0046] In one embodiment, when obtaining the M consecutive historical classification labels obtained by the latest classification of the tracking target, the specific process is to obtain the M consecutive historical classification labels obtained by the latest classification of the tracking target within the switching time threshold. Based on the switching time threshold, according to the M historical classification labels, it is determined that the label is frequently switched when the tracking target is classified. The specific implementation process is as follows: according to the M historical classification labels, the number of label switching is obtained; when the number of label switching is greater than the switching number threshold, it is determined that the label is frequently switched when the tracking target is classified. Among the M historical classification labels, if the mth historical label is different from the m+1th historical label, it is a label switch, and m is an integer greater than or equal to 1 and less than M.

[0047] In this embodiment, the switching time threshold is set according to actual conditions and needs, and the protection scope of this application is not limited to the specific value and setting method of the switching time threshold. Correspondingly, the switching number threshold is also set according to actual conditions and needs, and the protection scope of this application is also not limited to the specific value and setting method of the switching number threshold.

[0048] In a specific example, the switching time threshold is 2 minutes, and the switching number threshold is 3 times. Five historical classification labels are obtained within 2 minutes before the current moment, and the five historical classification labels are arranged in the order of acquisition time, namely, human, non-human, human, non-human, and human. Then, through the five historical classification labels, it can be obtained that the number of label switching is 4 times. Because the switching number threshold is 3 times, the number of label switching is greater than the switching number threshold, and it is determined that the label is frequently switched when classifying the tracking target.

[0049] In a specific embodiment, when determining whether the label is frequently switched when classifying the tracking target, at least one historical classification label within the switching time threshold can be obtained through the first counter and the first timer. Figure 2 As shown, the specific process in the classification process is as follows:

[0050] Step 201, initializing a first timer to zero, initializing a first counter to zero;

[0051] Step 202, when the category label is switched, the first timer starts timing;

[0052] Step 203, determining whether the timing of the first timer is less than a preset switching time threshold, if so, executing step 204, if not, executing step 201;

[0053] Step 204, the first counter is incremented by one;

[0054] Step 205, determining whether the count of the first counter reaches a preset switching number threshold, if so, executing step 206, if not, executing step 202;

[0055] Step 206, determining that labels frequently switch when tracking target classification.

[0056] When the label is frequently switched when determining the classification of the tracking target, the switching delay processing is performed on the tracking target.

[0057] In this embodiment, by setting the switching time threshold, it is avoided that only the number of switching times is counted, resulting in a large number of switching times in a long period of time, which does not belong to frequent switching, but is still confirmed as frequent switching.

[0058] In one embodiment, when obtaining the latest M continuous historical classification labels obtained by classifying the tracking target, the historical classification labels are saved in a preset label list. Specifically, the tracking target is classified in real time, and the classification labels obtained in real time each time are saved in the preset label list as historical classification labels; wherein the preset label list includes M historical classification labels. Based on the preset label list, according to the M historical classification labels, it is determined that the labels are frequently switched when classifying the tracking target. The specific implementation process is as follows: according to the M historical classification labels in the preset label list, it is determined that the labels are frequently switched when classifying the tracking target.

[0059] In this embodiment, the historical classification labels in the preset label list are arranged in the order of the time when the classification labels are obtained. For example, label 1 is obtained at 8:10, label 2 is obtained at 8:20, and label 3 is obtained at 8:30. Then, in the preset label list, the order of the saved labels is label 1, label 2, and label 3.

[0060] In one embodiment, the number of historical classification tags that can be saved in the preset tag list is limited and cannot exceed the preset storage threshold. Specifically, when the number of historical classification tags stored in the preset tag list reaches the preset storage threshold, when a new classification tag is obtained in real time, the historical classification tag that has been saved the longest in the preset tag list is deleted, and the new classification tag obtained in real time is saved as a historical classification tag in the preset tag list, and the preset tag list is updated to determine whether the tags are frequently switched next time. For example, when the historical classification tags saved in the preset tag list are Tag 1, Tag 2, and Tag 3, and the new classification tag obtained in real time is Tag 4, Tag 1 in the preset tag list is deleted, and Tag 4 is saved in the preset tag list. The updated preset tag list is Tag 2, Tag 3, and Tag 4.

[0061] Step 104: Obtain a target classification label, and perform switching delay processing on the tracked target using the target classification label.

[0062] In this embodiment, when the label of the tracking target switches frequently, the tracking target is delayed to determine a final target classification label, so as to avoid frequent label switching and reduce the accuracy of the classification result.

[0063] In one embodiment, the process of obtaining the target classification label is as follows: obtain N consecutive historical classification labels obtained from the latest classification of the tracking target, where N is an integer greater than 1; calculate the proportion of each historical classification label among the N historical classification labels; compare the proportion of each historical classification label with the preset proportion threshold corresponding to each historical classification label; and use the historical classification label whose proportion is greater than the preset proportion threshold as the target classification label.

[0064] In this embodiment, the preset ratio threshold corresponding to each historical classification label is pre-set according to actual conditions and needs, and the protection scope of this application is not limited to the specific value and setting method of the preset ratio threshold.

[0065] In a specific example, N is set to 15, and the classification labels include two labels: human and non-human. The preset ratio threshold corresponding to human is 70%, and the preset ratio threshold corresponding to non-human is 30%. When there are 12 human classification labels among the 15 historical classification labels, the proportion of human in the 15 historical classification labels is 80%. 80% is greater than 70%, so the target classification label is human.

[0066] In one embodiment, after obtaining the target classification label, the target tracking target is switched and delayed by the target classification label. The specific process is: keep the latest classification result of the tracking target as the target classification label unchanged until the holding time reaches the preset delay time, for example, the preset delay time is set to 5 minutes. The preset delay time is pre-set according to actual conditions and needs, and the protection scope of this application is not limited to the specific value and setting method of the preset delay time.

[0067] In one embodiment, when the holding time reaches the preset delay time, the tracking target is reclassified until it is determined that the label is frequently switched when classifying the tracking target based on the M historical classification labels. In this way, the process of determining whether the frequent switching forms a cyclic process, thereby avoiding the process of frequent switching but being unable to be processed again after a delay processing, and further improving the accuracy of target classification.

[0068] Based on the above embodiment, the historical classification labels whose proportion is greater than the preset ratio threshold are used as target classification labels, and the target classification labels are provided to the user. In this way, the information obtained by the user is no longer the classification labels that jump back and forth, but the obtained target classification labels. In particular, the effect of using classification labels in other processing processes is particularly obvious. For example, when the classification labels are used to determine whether the automatic feeding device of the pet is automatically turned on, when the labels are frequently switched, the automatic feeding device will be constantly turned on and off due to the uncertainty of the obtained classification labels, and the automatic feeding device will be set to cause the automatic opening function of the automatic feeding device to be disordered, or the automatic opening device of the automatic feeding device will be damaged. Through this method, when the labels are frequently switched, the determined target classification labels can make the automatic feeding device not frequently turned on or off, thereby avoiding damage to the automatic feeding device. Thus, more intelligent, humane and safe services are provided to users.

[0069] In one embodiment, when the labels are not frequently switched during the classification of the tracking target, the final label of the tracking target can be determined by the proportion of historical classification labels and the preset proportion threshold corresponding to each historical classification label. Specifically, the final label determination process can be completed by a maximum of N historical classification labels.

[0070] For example, N takes a value of 15, and the classification labels include two labels: human and sub-human. The preset ratio threshold corresponding to human is 70%, and the preset ratio threshold corresponding to non-human is 30%. At the beginning of classification, the number of historical classification labels is less than 15. When it is less than 15 times at the beginning, the classification label obtained in the first classification is human, and the proportion of human is 100%, and the final label is human; continue to classify, the classification label obtained in the second classification is human, and the proportion of human is 100%, and the final label is human; the classification label obtained in the third classification is non-human, and the proportion of human is 66.67%, which is less than 70%, and the proportion of non-human is 33.33%, which is greater than 30%, and the final label is non-human; the classification label obtained in the fourth classification is non-human, and the proportion of human is 75%, and the final label is human; the classification label obtained in the fifth classification is human, and the classification label obtained in the classification is non-human, and the proportion of human is 80%, and the final label is human, and so on.

[0071] When the historical classification labels are greater than 15, during the 16th classification, the historical classification labels obtained from the first classification are no longer used. The classification labels obtained from the 16th classification are added, and the proportion (that is, the proportion of the total from the 2nd to the 16th classification) is determined. The classification labels obtained from the 17th classification replace the classification labels obtained from the second classification. The results of subsequent classifications are processed in sequence to complete the determination of the final labels.

[0072] In one embodiment, after obtaining the target classification label, based on the method of the above embodiment, the classification label of the tracking target is continuously monitored to see whether frequent switching occurs again. When frequent label switching occurs again, the target classification label is re-determined to realize the dynamic process of label recognition and delay processing.

[0073] In one embodiment, if Figure 3 As shown, when the target classification method is implemented, the initialization threshold and variable counter are set. The initialization threshold includes the frequent change number threshold and the same label duration threshold when the target is determined to be a frequently changed target. The variable counter includes a frequently switched target to keep the same label timer and a frequent change number counter. The specific process of target classification is as follows:

[0074] Step 301, determining a tracking target;

[0075] Step 302, when the classification label of the tracking target changes once, the frequent change count counter is incremented by one;

[0076] Step 303, determine whether the frequent change count counter is greater than the frequent change count threshold, if so, execute step 304, if not, execute step 301;

[0077] Step 304, by frequently switching the target and maintaining the same label timer, obtain the duration of the classification label unchanged;

[0078] Step 305, when the classification label changes, determine whether the duration is greater than the first time threshold, if so, execute step 306, if not, execute step 304;

[0079] Step 306 , determining the classification label that remains unchanged within the duration as the target classification label, initializing the frequent change count counter to zero, and executing step 301 .

[0080] In one embodiment, if Figure 4 As shown in Figure 2, the implementation process of the target classification method is as follows:

[0081] Step 401, initializing threshold setting.

[0082] Specifically, the initialization threshold and variable timer are set, and the initialization threshold includes the frequent change number threshold, the time threshold between two adjacent label changes, and the same label duration threshold when the target is determined to be a frequently changed target. The above thresholds are all set in advance according to actual conditions and needs.

[0083] Among them, the time threshold between two adjacent label changes can be achieved by setting a frame threshold. Specifically, the time corresponding to a frame is pre-defined, and the frame threshold represents the number of frames. For example, 0.1 seconds is used as a frame in the detection process, and the frame threshold is 20 frames. Correspondingly, the time threshold between two adjacent label changes is 2 seconds. In this embodiment, the specific rules and values ​​for setting the frame threshold are determined according to actual conditions and needs, and the protection scope of this application is not limited to the specific rules and values ​​for setting the frame threshold.

[0084] The variable timer includes a timer for keeping the same label when switching targets frequently, a timer for the actual label when keeping the same label, a timer for two consecutive label changes, and a timer for the number of frequent changes. The above timers are automatically started during the target classification process to complete the continuous timing function.

[0085] Step 402, determine whether the target is locked, if so, execute step 403, if not, execute step 4022.

[0086] Step 4022: No switching delay processing is performed.

[0087] Specifically, target locking refers to the process of continuously classifying the target after it is determined to be a target for tracking. When the target cannot be locked, that is, the target cannot be tracked, the switching delay process is not performed.

[0088] Step 403 , determine whether the tag of the tracking target is frequently switched, if so, execute step 404 , if not, execute step 4032 .

[0089] Step 4032, identifying again whether the tag of the tracking target is frequently switched.

[0090] Step 404, maintaining the same tag timing delay processing.

[0091] After determining that the tracking target label switches frequently, the delay processing process begins.

[0092] Step 405, the actual classification label maintains the same label timing.

[0093] After entering the delayed processing process, the actual classification label is started to keep the same label timing.

[0094] Step 406 , determining whether the tracking target triggers classification in the current frame, if so, executing step 407 , if not, executing step 411 .

[0095] In this embodiment, the time corresponding to one frame is predefined, and whether to trigger the classification of the tracking target is determined in each frame in units of frames, so as to continuously track the tracking target and start the classification process under appropriate conditions during the tracking process.

[0096] Step 407, determine whether the current classification label is the same as the previous classification label, if so, execute step 408, if not, execute step 409.

[0097] In step 408 , the classification label remains unchanged, the counter of the same label is kept accumulated, and step 410 is executed.

[0098] Step 409, the classification label is changed, the actual classification label keeps the same label timer and resets to zero, and then executes step 410.

[0099] Step 410, the tag does not change frequently to confirm the detection.

[0100] Enter the detection process where there is no frequent change of detection labels.

[0101] Step 411, determine whether the delay time is greater than the delay threshold, if so, execute step 412, if not, execute step 423.

[0102] The delay time is obtained by counting the delay timer, and the delay threshold is a preset threshold.

[0103] Step 412, update the classification label and reset the delay timer to zero.

[0104] Step 413, fix the target classification label.

[0105] That is, determining the target classification label in the above embodiment.

[0106] In this embodiment, based on Figure 4When the tracking target is identified again, whether the label is frequently switched, such as Figure 5 As shown, the specific process is as follows:

[0107] Step 501, timing of two consecutive label changes.

[0108] When it is identified again whether the label of the tracking target is frequently switched, a timer for two consecutive label changes is started to time the time interval between the two consecutive label changes.

[0109] Step 502 , determining whether the time between two adjacent label changes is greater than the time threshold between two adjacent label changes, if so, executing step 503 , if not, executing step 504 .

[0110] Step 503: The timers for two consecutive label changes and the timer for the number of frequent changes are reset to zero.

[0111] Step 504 , determine whether the tracking target triggers classification in the current frame, if so, execute step 505 , if not, execute step 508 .

[0112] Step 505, determine whether the current classification label is the same as the previous classification label, if so, execute step 506, if not, execute step 508.

[0113] Step 506, the classification label is changed, and after confirming that the change is a frequent change, the frequent change count timer is accumulated.

[0114] Step 507: The timers of two consecutive label changes are reset to zero.

[0115] Step 508, the tag frequently changes the target to confirm the detection.

[0116] See Figure 4 Frequent changes of labels in the target confirm the detection process.

[0117] Step 509 , determine whether the frequent change count timer is greater than the frequent change count threshold, if so, execute step 510 , if not, execute step 5091 .

[0118] Step 5091, continue to classify the tracking target.

[0119] Step 510: confirm that the tag frequently changes the target identification bit.

[0120] After confirming the identification bit, it means that the tracking target is confirmed to have frequent tag changes.

[0121] Step 511, confirm the target classification label to be delayed, that is, to remain unchanged.

[0122] See Figure 4 The specific process of confirming the target classification label in .

[0123] Step 512, the frequently switched target keeping the same label timer and the frequently changed times timer are reset to zero.

[0124] The target classification method provided by the present application determines the tracking target, and then determines when the tracking target frequently switches labels during classification based on the M continuous historical classification labels obtained from the latest classification of the tracking target, and then obtains the target classification label, and performs switching delay processing on the tracking target through the target classification label. Through this method, when tracking and classifying the tracking target, the continuous and frequent switching of the classification label is avoided. When frequent switching occurs, delay processing is immediately performed to obtain the target classification label after delay processing, so as to achieve smooth optimization of the classification process and thereby improve the classification accuracy. Avoid poor usage effects caused by frequent label switching and improve the user experience.

[0125] In the process of implementing the target classification method, when millimeter-wave radar technology is used, based on the advantages of millimeter-wave radar technology, the signal used for target classification is more precise, further improving the accuracy of target classification.

[0126] Based on the same concept, an object classification device is provided in the embodiment of the present application. The specific implementation of the device can be found in the description of the method embodiment part, and the repeated parts will not be repeated. Figure 6 As shown, the device mainly includes:

[0127] A first determination module 601 is used to determine a tracking target;

[0128] An acquisition module 602 is used to acquire M continuous historical classification labels obtained by the latest classification of the tracking target, where M is an integer greater than 1;

[0129] A second determination module 603 is used to determine, based on the M historical classification labels, that labels are frequently switched when classifying the tracking target;

[0130] The delay processing module 604 is used to obtain a target classification label and perform a switching delay process on the tracking target according to the target classification label.

[0131] Based on the same concept, an electronic device is also provided in the embodiment of the present application, such as Figure 7As shown, the electronic device mainly includes: a processor 701, a memory 702 and a communication bus 703, wherein the processor 701 and the memory 702 communicate with each other through the communication bus 703. The memory 702 stores a program executable by the processor 701, and the processor 701 executes the program stored in the memory 702 to implement the following steps: determining a tracking target; obtaining at least one historical classification label of the tracking target; determining that the label is frequently switched when classifying the tracking target according to the at least one historical classification label; performing a switching delay process on the tracking target with the label frequently switched, and obtaining the target classification label after the delay process.

[0132] The communication bus 703 mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 703 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0133] The memory 702 may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Alternatively, the memory may also be at least one storage device located away from the processor 701.

[0134] The above-mentioned processor 701 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., and can 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 gates or transistor logic devices, discrete hardware components.

[0135] In another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed on a computer, the computer executes the target classification method described in the above embodiment.

[0136] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instruction is loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instruction can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instruction is transmitted from a website site, a computer, a server or a data center by wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, a data center, etc. that contains one or more available media integration. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape, etc.), an optical medium (such as a DVD) or a semiconductor medium (such as a solid-state hard disk), etc.

[0137] It should be noted that, in this article, relational terms such as "first" and "second" are only used 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.

[0138] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A target classification method, characterized in that: include: Determine the tracking target; Obtaining M continuous historical classification labels obtained by the latest classification of the tracking target, where M is an integer greater than 1; Determining, based on the M historical classification labels, that labels are frequently switched when classifying the tracking target; Acquire a target classification label, and perform a switching delay process on the tracking target by using the target classification label; The method of obtaining a target classification label includes: obtaining N consecutive historical classification labels obtained from the latest classification of the tracking target, wherein N is an integer greater than 1; calculating the proportion of each historical classification label among the N historical classification labels; comparing the proportion of each historical classification label with a preset proportion threshold corresponding to each historical classification label; and using the historical classification label whose proportion is greater than the preset proportion threshold as the target classification label.

2. The target classification method according to claim 1, characterized in that: The obtaining of the M continuous historical classification labels obtained by the latest classification of the tracking target includes: The M continuous historical classification labels obtained by the latest classification of the tracking target within the switching time threshold; The determining, based on the M historical classification labels, that labels are frequently switched when classifying the tracking target comprises: According to the M historical classification labels, obtain the number of label switching times; When the number of label switching times is greater than the switching number threshold, it is determined that labels are frequently switched when classifying the tracking target.

3. The target classification method according to claim 1, characterized in that: The determining of the tracking target comprises: Acquire the millimeter-wave radar echo signal of the detection target; Analyze the echo signal to obtain the current distance of the target from the millimeter wave radar; When the current distance is less than a first distance threshold, the target is determined to be the tracking target.

4. The target classification method according to claim 3, characterized in that: The obtaining of the M continuous historical classification labels obtained by the latest classification of the tracking target includes: Tracking the trajectory of the tracking target by using the millimeter wave radar; Obtain the echo signal of the millimeter-wave radar during trajectory tracking; By using a machine learning classification algorithm, the echo signal of the millimeter wave radar during the trajectory tracking process is analyzed to continuously classify the tracking target; Obtain M continuous historical classification labels obtained by the latest classification of the tracking target.

5. The target classification method according to claim 1, characterized in that: The performing switching delay processing on the tracking target by using the target classification label includes: The latest classification result of the tracked target is maintained as the target classification label unchanged until the holding time reaches a preset delay time.

6. The target classification method according to claim 5, characterized in that: When the holding time reaches the preset delay time, the method further includes: The tracking target is reclassified until it is determined again based on the M historical classification labels that labels are frequently switched when classifying the tracking target.

7. A target classification device, characterized in that: include: A first determination module, used to determine a tracking target; An acquisition module, used to acquire M continuous historical classification labels obtained by the latest classification of the tracking target, where M is an integer greater than 1; A second determination module is used to determine, based on the M historical classification labels, that labels are frequently switched when classifying the tracking target; A delay processing module, used for obtaining a target classification label and performing a switching delay processing on the tracking target according to the target classification label; The delay processing module is also used to: obtain N consecutive historical classification labels obtained from the latest classification of the tracking target, wherein N is an integer greater than 1; calculate the proportion of each historical classification label among the N historical classification labels; compare the proportion of each historical classification label with a preset proportion threshold corresponding to each historical classification label; and use the historical classification label whose proportion is greater than the preset proportion threshold as the target classification label.

8. An electronic device, characterized in that: include: A processor, a memory and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory is used to store computer programs; The processor is used to execute the program stored in the memory to implement the target classification method described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the target classification method according to any one of claims 1 to 6 is implemented.

Images