Target tracking method, storage medium and electronic device

Abstract

This application provides a target tracking method, a storage medium, and an electronic device. The target tracking method includes: acquiring multiple images; determining a first ground truth target sub-image from a starting image of the multiple images; performing at least one transformation on the first ground truth target sub-image to obtain at least one transformed first ground truth target sub-image; constructing multiple training sets based on the first ground truth target sub-image and the at least one transformed first ground truth target sub-image; training multiple filters based on the multiple training sets; and tracking a target object in the multiple images based on the multiple filters. The target tracking method has good tracking performance.

Description

Technical Field

[0001] This application belongs to the field of image processing technology, and relates to a tracking method, particularly a target tracking method, storage medium, and electronic device. Background Technology

[0002] Target tracking typically refers to predicting the subsequent position of an object in a video sequence, given the positional information of an object in one frame. Target tracking technology has wide applications in video surveillance, autonomous driving, human-computer interaction, and other scenarios. Summary of the Invention

[0003] In a first aspect, embodiments of this application provide a target tracking method. The target tracking method includes: acquiring multiple collected images; determining a first ground truth target sub-image from a starting image of the multiple images; performing at least one transformation on the first ground truth target sub-image to obtain at least one transformed first ground truth target sub-image; constructing multiple training sets based on the first ground truth target sub-image and the at least one transformed first ground truth target sub-image; training multiple filters based on the multiple training sets to obtain multiple filters; and tracking target objects in the multiple images based on the multiple filters.

[0004] In one implementation of the first aspect, performing at least one transformation on the first truth target subgraph to obtain at least one transformed first truth target subgraph includes: performing a shrinking transformation on the first truth target subgraph to obtain a second truth target subgraph, and / or performing a magnifying transformation on the first truth target subgraph to obtain a third truth target subgraph.

[0005] In one implementation of the first aspect, the target tracking method further includes: constructing a first training set based on the first ground truth target subgraph, constructing a second training set based on the second ground truth target subgraph, and constructing a third training set based on the third ground truth target subgraph; training a first filter based on the first training set, training a second filter based on the second training set, and training a third filter based on the third training set.

[0006] In one implementation of the first aspect, the target tracking method further includes: obtaining a first sub-image from the current frame image based on the position of the target object in the previous frame image, wherein the first sub-image and the first ground truth target sub-image have the same size; tracking the target object based on the first sub-image and the first filter; determining that tracking has failed, performing a first offset on the position to obtain at least one first offset position, wherein the distance of the first offset is determined by the first ground truth target sub-image; obtaining at least one second sub-image from the current frame image based on the at least one first offset position, wherein the size of the at least one second sub-image is the same as the size of the first ground truth target sub-image; and tracking the target object based on the at least one second sub-image and the first filter.

[0007] In one implementation of the first aspect, the target tracking method further includes: determining that tracking has failed; obtaining a third sub-image from the current frame image based on the position, wherein the third sub-image and the second ground truth target sub-image have the same size; tracking the target object based on the third sub-image and the second filter; determining that tracking has failed, performing a second offset on the position to obtain at least one second offset position, wherein the distance of the second offset is determined by the second ground truth target sub-image; obtaining at least one fourth sub-image from the current frame image based on the at least one second offset position, wherein the size of the at least one fourth sub-image is the same as the size of the second ground truth target sub-image; and tracking the target object based on the at least one fourth sub-image and the second filter.

[0008] In one implementation of the first aspect, the target tracking method further includes: determining that tracking has failed; obtaining a fifth sub-image from the current frame image based on the position, wherein the fifth sub-image and the third ground truth target sub-image of the target object have the same size; tracking the target object based on the fifth sub-image and the third filter; determining that tracking has failed, performing a third offset on the position to obtain at least one third offset position, wherein the distance of the third offset is determined by the third ground truth target sub-image; obtaining at least one sixth sub-image from the current frame image based on the at least one third offset position, wherein the size of the at least one sixth sub-image is the same as the size of the third ground truth target sub-image; and tracking the target object based on the at least one sixth sub-image and the third filter.

[0009] In one implementation of the first aspect, determining tracking failure includes: obtaining the sidelobe ratio of the corresponding subgraph according to the filter, and determining tracking failure based on the sidelobe ratio of the subgraph being less than a sidelobe ratio threshold.

[0010] In one implementation of the first aspect, obtaining the sidelobe ratio of the corresponding sub-image based on the filter includes: obtaining the discrete Fourier transform result of the corresponding sub-image; performing an inverse discrete Fourier transform on the product of the filter and the discrete Fourier transform result to obtain an intermediate image; and obtaining the sidelobe ratio of the corresponding sub-image based on the peak value, mean, and standard deviation of the intermediate image.

[0011] In a second aspect, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the target tracking method described in any one of the first aspects of this application.

[0012] Thirdly, embodiments of this application provide an electronic device, the electronic device comprising: a memory storing a computer program; and a processor communicatively connected to the memory, which executes the target tracking method described in any one of the first aspects of this application when the computer program is invoked. Attached Figure Description

[0013] Figure 1 The diagram shown is a structural schematic of an electronic device according to an embodiment of this application.

[0014] Figure 2 The flowchart shown is a target tracking method according to an embodiment of this application.

[0015] Figure 3A The diagram shown is a flowchart of a filter training method in one embodiment of this application.

[0016] Figure 3B The image shown is an example of the starting image in one embodiment of this application.

[0017] Figure 3C The diagram shown is an example of a first truth target subgraph in one embodiment of this application.

[0018] Figure 3D and Figure 3E The images shown are example diagrams of training images and ground truth labels in one embodiment of this application.

[0019] Figure 4 The flowchart shown is a process for tracking a target object in one embodiment of this application.

[0020] Figure 5 The flowchart shown is a process for tracking a target object in one embodiment of this application.

[0021] Figure 6 The flowchart shown is a process for tracking a target object in one embodiment of this application.

[0022] Figure 7 The flowchart shown is a process for obtaining the sidelobe ratio of the first subgraph in one embodiment of this application.

[0023] Figure 8 The diagram shown is a schematic representation of a position offset in one embodiment of this application. Detailed Implementation

[0024] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0025] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0026] The following embodiments of this application provide a target tracking method that can be applied to electronic devices. The electronic devices described in these embodiments may include mobile phones, tablets, wearable devices, in-vehicle devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), and other terminal devices, but are not limited thereto. This application does not impose any restrictions on the specific type of electronic device.

[0027] For example, the electronic device may be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, a computer, a laptop computer, a handheld communication device, a handheld computing device, and / or other devices for communicating over a wireless system, as well as next-generation communication systems, such as mobile terminals in 5G networks, mobile terminals in future evolved Public Land Mobile Networks (PLMNs), or mobile terminals in future evolved Non-terrestrial Networks (NTNs).

[0028] As an example and not a limitation, when the electronic device is a wearable device, it can also be a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as gloves and watches equipped with near-field communication modules. Wearable devices are portable devices worn directly on the body or integrated into a user's clothing or accessories. By attaching to the user and using a pre-bound electronic card, they perform operations such as payment and authentication. Wearable devices are not merely hardware devices; they achieve powerful functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices include those with comprehensive functions, large size, and the ability to perform complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those focused on a specific application function and requiring the use of other devices such as smartphones, such as various smartwatches and smart bracelets with displays.

[0029] The following will describe in detail, with reference to the accompanying drawings, the principle and implementation of a target tracking method provided in this application, so that those skilled in the art can understand the target tracking method provided in this application without creative effort.

[0030] Figure 1 The diagram shown illustrates the structure of an electronic device 1 used in the target tracking method provided in this embodiment of the application. Figure 1 As shown, the electronic device 1 includes an image acquisition device 11, an image signal processor (ISP) 12, at least one general-purpose processor 13, a display 14, and a memory 15.

[0031] The image acquisition device 11 may include components such as a camera 111 and an image sensor 112 for acquiring images or videos of the target scene. The camera 111 may be positioned at the front (front-facing camera) or rear (rear-facing camera) of the main body of the electronic device 1. The image sensor 112 is a photosensitive element, such as a complementary metal-oxide-semiconductor (CMOS) element or a charge-coupled device (CCD). The function of the image sensor 112 is to capture the optical image acquired by the camera 111 and convert it into an electrical signal usable by the image signal processor 12.

[0032] The main function of the image signal processor 12 is to process the signal output by the front-end image sensor 112. In this embodiment, the image signal processor 12 may include, but is not limited to, one or more of the following processing algorithms: Automatic Exposure Control (AEC), Automatic Gain Control (AGC), color correction, lens correction, noise removal / reduction, bad pixel removal, linear correction, color interpolation, image downsampling, level compensation, etc.

[0033] The general-purpose processor 13 can be any type of device capable of processing electronic instructions. In this embodiment, the electronic device 1 may include one or more general-purpose processors 13, such as one or both of a central processing unit (CPU) 131 and a neural-network processing unit (NPU) 132. Furthermore, it may include one or more of a graphics processing unit (GPU), microprocessor, microcontroller, main processor, controller, and ASIC (Application Specific Integrated Circuit). The general-purpose processor 13 is configured to execute various types of digital storage instructions, such as software or firmware programs stored in memory 15, enabling the electronic device 1 to provide various services. For example, the processor 11 can execute programs or process data to perform at least a portion of the target tracking method provided in this embodiment.

[0034] The main function of the central processing unit 131 is to parse computer instructions and process data in computer software, realize the overall control of electronic device 1, and control and allocate all hardware resources of electronic device 1 (such as storage resources, communication resources, I / O interfaces, etc.).

[0035] Neural network processors are a general term for new types of processors based on neural network algorithms and acceleration. They are specifically designed for artificial intelligence and are used to accelerate neural network operations, solving the problem of low efficiency of traditional chips in neural network operations.

[0036] It should be noted that the name of the neural network processor does not constitute a limitation of this application. For example, in other application scenarios, the neural network processor can also be modified or replaced by other processors with similar functions, such as a tensor processing unit (TPU), a deep learning processing unit (DPU), and so on.

[0037] The display 14 may specifically include a display screen (display panel). In some implementations, the display panel may be configured using a liquid crystal display (LCD), an organic light-emitting diode (OLED), or other similar forms. The display device may also be a touch panel (touchscreen, touch screen), which may include a display screen and a touch-sensitive surface. When the touch-sensitive surface detects a touch operation on or near it, it transmits the information to the central processing unit 131 to determine the type of touch event. Subsequently, the central processing unit 131 provides corresponding visual output on the display device based on the type of touch event.

[0038] The memory 15 may include volatile memory, such as random access memory (RAM) or cache. The memory 15 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD). The memory 15 can be used to store data such as images acquired by the image acquisition device 11. The memory 15 can also be used to store program instructions for the processor to call and execute the target tracking method described in this application.

[0039] Figure 2 The flowchart shown is a target tracking method provided in an embodiment of this application. For example... Figure 2 As shown, the target tracking method provided in this application embodiment includes steps S21 to S25.

[0040] S21, acquire multiple images. These multiple images may be, for example, multiple frames of video captured by the image acquisition device 11, or images acquired and transmitted by other electronic devices.

[0041] S22, determine a first ground truth target sub-image from a starting image of multiple images. The first ground truth target sub-image contains the target object to be tracked. For example, in a face tracking application scenario, the first ground truth target sub-image may be an image containing a human face.

[0042] In some implementations, a detection algorithm can be used to process the initial image of the multiple images to obtain a detection box, and the content of the detection box can be cropped to obtain the first ground truth target sub-image. The detection algorithm can be a neural network such as FaceNet or MTCNN (Multi-task Cascaded Convolutional Networks), which is not a limitation of this application.

[0043] In other implementations, the starting image of the multiple images can be manually selected to obtain the truth box, and the image in the truth box can be cropped to obtain the first truth target sub-image.

[0044] S23, perform at least one transformation on the first truth target subgraph to obtain at least one transformed first truth target subgraph.

[0045] S24, construct multiple training sets based on the first ground truth target sub-graph and at least one modified first ground truth target sub-graph respectively, and train multiple filters based on the multiple training sets.

[0046] S25 tracks target objects in multiple images based on multiple filters.

[0047] In one embodiment of this application, performing at least one transformation on a first truth target subgraph to obtain at least one transformed first truth target subgraph includes: reducing the size of the first truth target subgraph to obtain a second truth target subgraph, and / or enlarging the first truth target subgraph to obtain a third truth target subgraph. The reduction and enlargement ratios can be preset values ​​or configured according to actual needs.

[0048] In some implementations, the center of the first truth target subgraph can be used as a reference, and the size of the first truth target subgraph can be reduced to size1 = a × size0 to obtain the second truth target subgraph, where a is a positive number less than 1, and size0 is the size of the first truth target subgraph, for example, a is 0.7.

[0049] In some other implementations, the center of the first truth target subgraph can be used as a reference, and the size of the first truth target subgraph can be enlarged to size2 = b × size0 to obtain the third truth target subgraph, where b is a positive number greater than 1, and size0 is the size of the first truth target subgraph, for example, b is 1.3.

[0050] Please see Figure 3A In one embodiment of this application, the target tracking method may further include the following steps S31 and S32.

[0051] S31, construct a first training set based on the first ground truth target subgraph, construct a second training set based on the second ground truth target subgraph, and construct a third training set based on the third ground truth target subgraph.

[0052] S32, a first filter is obtained by training based on the first training set, a second filter is obtained by training based on the second training set, and a third filter is obtained by training based on the third training set.

[0053] Taking the training of a first filter using a first training set as an example, in this embodiment of the application, a series of training images and the ground value labels of each training image can be obtained by performing a series of rotation and translation transformations on the first ground value target sub-image, which serves as the first training set. For example, Figure 3B The image shown is an example of the starting image in an embodiment of this application. Figure 3C The image shown is an example of the first truth target subgraph in this application embodiment. Figure 3C The first truth target subplot shown is obtained by... Figure 3B The image is obtained by processing the initial image. The processing method can be either the detection algorithm described in the embodiments of this application or manual processing. Figure 3D and Figure 3E These are example images of a series of training images and ground truth labels shown in the embodiments of this application. Figure 3E The bright spots represent the center position of the target object in the image. Based on the training image and its ground truth labels, the first filter H0 can be obtained through Equation 1 in this embodiment.

[0054]

[0055] Where N represents the number of training images, in Figure 3E In the example shown, N = 6. Fi G represents the Fourier transform result of the i-th training image. i This represents the Fourier transform result of the i-th truth label.

[0056] Please see Figure 4 In one embodiment of this application, the target tracking method may further include the following steps S41 to S45.

[0057] S41, obtain a first sub-image from the current frame image based on the position of the target object in the previous frame image. The first sub-image and the first ground truth target sub-image have the same size, size0. For example, at the stated position, a sub-image of size 0 can be obtained from the current frame image as the first sub-image. In some examples, the position can be represented by pixel coordinates, typically including the pixel coordinates of the upper left and lower right corners. The position of the target object in the image can be determined based on these two pixel coordinates.

[0058] S42, Track the target object based on the first sub-image and the first filter.

[0059] S43, if tracking fails, perform a first offset on the position to obtain at least one first offset position. The distance of the first offset is determined by a first ground truth target sub-image. In some implementations, the distance of the first offset can be half the distance of the first ground truth target sub-image, but this application is not limited to this.

[0060] S44, at least one second sub-image is obtained from the current frame image based on at least one first offset position, wherein the size of the at least one second sub-image is the same as the size of the first ground truth target sub-image, which is size0. For example, a sub-image of size0 can be obtained from the current frame image at the first offset position as the second sub-image.

[0061] S45, the target object is tracked based on at least one second subgraph and a first filter.

[0062] In this embodiment, steps S41 and S42 constitute the first tracking stage, and steps S43 to S45 constitute the second tracking stage. Therefore, the target tracking method provided in this embodiment includes at least two tracking stages, entering the second tracking stage after the first tracking stage fails. Compared to traditional tracking methods, the target tracking method provided in this embodiment exhibits better tracking performance when the target object undergoes significant displacement and scale changes. Furthermore, this target tracking method only activates the second tracking stage if the first tracking stage fails; it does not activate the second tracking stage if the first tracking stage succeeds. Thus, it can minimize computational overhead while improving tracking performance.

[0063] Please see Figure 5 In one embodiment of this application, the target tracking method may further include the following steps S51 to S56.

[0064] S51, tracking failure is determined. Specifically, tracking of the target object based on at least one second subgraph and a first filter is determined to have failed.

[0065] S52, obtain a third sub-image from the current frame image according to the position, wherein the third sub-image and the second ground truth target sub-image have the same size.

[0066] S53, track the target object based on the third subgraph and the second filter.

[0067] S54, if tracking fails, perform a second offset on the position to obtain at least one second offset position, wherein the distance of the second offset is determined by the second ground truth target sub-image. In some implementations, the distance of the second offset can be 1 / 2 of the second ground truth target sub-image, but this application is not limited thereto.

[0068] S55, at least one fourth sub-image is obtained from the current frame image based on at least one second offset position, wherein the size of the at least one fourth sub-image is the same as the size of the second ground truth target sub-image.

[0069] S56, the target object is tracked based on at least one fourth subgraph and a second filter.

[0070] In this embodiment, steps S51 to S53 constitute the third tracking stage, and steps S54 to S56 constitute the fourth tracking stage. Therefore, in this embodiment, when the second tracking stage fails, the system proceeds to the third tracking stage; when the third tracking stage fails, the system proceeds to the fourth tracking stage.

[0071] Please see Figure 6 In one embodiment of this application, the target tracking method may further include the following steps S61 to S66.

[0072] S61, tracking failure is determined. Specifically, tracking of the target object based on at least one fourth subgraph and a second filter is determined to have failed.

[0073] S62, obtain a fifth sub-image from the current frame image according to the position, wherein the fifth sub-image and the third ground truth target sub-image of the target object have the same size.

[0074] S63 tracks the target object based on the fifth subgraph and the third filter.

[0075] S64, if tracking fails, perform a third offset on the position to obtain at least one third offset position, wherein the distance of the third offset is determined by the third ground truth target sub-image. In some implementations, the distance of the third offset can be 1 / 2 of the third ground truth target sub-image, but this application is not limited thereto.

[0076] S65, at least one sixth sub-image is obtained from the current frame image based on at least one third offset position, wherein the size of the at least one sixth sub-image is the same as the size of the third ground truth target sub-image.

[0077] S66 tracks the target object based on at least one sixth subgraph and a third filter.

[0078] In this embodiment, steps S61 to S63 constitute the fifth tracking stage, and steps S64 to S66 constitute the sixth tracking stage. Therefore, in this embodiment, when the fourth tracking stage fails, the system proceeds to the fifth tracking stage; when the fifth tracking stage fails, the system proceeds to the sixth tracking stage.

[0079] It should be noted that the target tracking method provided in the embodiments of this application may also include a seventh tracking stage, an eighth tracking stage, a ninth tracking stage, etc. The difference between the different tracking stages lies in the different filters used and the different sizes of the sub-images. The number of specific tracking stages is not limited in this application.

[0080] In one embodiment of this application, determining tracking failure includes: obtaining the sidelobe ratio of the corresponding subgraph based on the filter, and determining tracking failure based on the fact that the sidelobe ratio of the subgraph is less than a sidelobe ratio threshold. The sidelobe ratio threshold can be configured according to actual needs.

[0081] Please see Figure 7 This is a flowchart illustrating the process of obtaining the sidelobe ratio of the first sub-graph based on the filter in an embodiment of this application. For example... Figure 7 As shown, in this embodiment of the application, obtaining the sidelobe ratio of the corresponding subgraph according to the filter includes the following steps S71 to S73.

[0082] S71, obtain the discrete Fourier transform result of the first subgraph.

[0083] S72, perform an inverse discrete Fourier transform on the product of the filter and the discrete Fourier transform result of the first sub-image to obtain the intermediate image g. Specifically, in step S72, the intermediate image g can be obtained using the following equation 2.

[0084] g = IDFT(FH0), Equation 2.

[0085] Where IDFT represents the inverse discrete Fourier transform, g represents the intermediate image, and g maxLet be the peak value of g, corresponding to the brightest point in the intermediate image g. In the intermediate image g, remove the region with dimensions M pixels near the brightest point. The mean and standard deviation of the remaining region are μ and σ, respectively, where M is a positive integer, for example, 11. H0 represents the first filter, and F is the discrete Fourier transform result of the first sub-image.

[0086] S73, based on the peak value g of the intermediate image max The sidelobe ratio of the first subplot is obtained from the mean μ and the standard deviation σ. Specifically, in step S73, the sidelobe ratio PSR of the first subplot can be obtained using the following Equation 3.

[0087]

[0088] In some implementations, if the sidelobe ratio of the first subgraph is greater than or equal to a sidelobe ratio threshold, then tracking is considered successful. max The position is the center position of the target object.

[0089] It should be noted that the sidelobe ratio of the second and third subgraphs can be achieved using a similar method, which will not be elaborated on here.

[0090] In one embodiment of this application, offsetting the position to obtain at least one first offset position includes: offsetting the position upwards, downwards, to the left, and to the right to obtain four first offset positions. For example, Figure 8 The diagram shows an offset illustration in this application embodiment. The dashed box in the diagram represents the position of the target object in the previous frame image, and the solid box represents the first offset position after offset.

[0091] It should be noted that the methods for obtaining the second and third offset positions are similar to those for obtaining the first offset position, and will not be elaborated on here.

[0092] This application also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the target tracking method described in any embodiment of this application.

[0093] In the embodiments of this application, any combination of one or more storage media can be used. The storage medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0094] This application also provides an electronic device. The electronic device includes: a memory storing a computer program; and a processor communicatively connected to the memory, which executes the target tracking method of any embodiment of this application when the computer program is invoked.

[0095] In some implementations, the electronic device may also include a display for showing a graphical user interface related to the target tracking method.

[0096] In summary, the target tracking method provided in this application adopts a multi-stage tracking approach, allowing subsequent stages to continue tracking even if the previous stage fails. Therefore, even if the target object undergoes significant displacement and scale changes, the target tracking method provided in this application can still achieve good tracking results. Furthermore, the target tracking method provided in this application can also employ a multi-stage, multi-filter approach for target tracking, and this approach is only activated when the tracking quality is poor, thus minimizing computational overhead while improving tracking performance. Therefore, this application effectively overcomes various shortcomings of the prior art and has high industrial applicability.

[0097] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A target tracking method, characterized in that, include: Acquire multiple images; A first ground truth target sub-image is determined from the starting image of the plurality of images, the first ground truth target sub-image containing the target object to be tracked; The first ground truth target sub-image is subjected to at least one size transformation to obtain at least one size-transformed first ground truth target sub-image; multiple training sets are constructed based on the training images obtained by rotation and translation transformation of the first ground truth target sub-image and the training images obtained by rotation and translation transformation of the at least one size-transformed first ground truth target sub-image, and multiple filters are trained based on the multiple training sets; the size of the ground truth target sub-images corresponding to different filters is different. The target object in the multiple images is tracked based on the multiple filters. The tracking includes at least two tracking stages. If the first tracking stage fails, a second tracking stage is entered. The first and second tracking stages use a first filter. Entering the second tracking stage after the first tracking stage fails includes: obtaining a first sub-image from the current frame image based on the target object's position in the previous frame image, wherein the first sub-image and the first ground truth target sub-image have the same size; tracking the target object in the first tracking stage based on the first sub-image and the first filter; determining that the first tracking stage has failed, performing a first offset on the position to obtain at least one first offset position, wherein the distance of the first offset is determined by the size of the first ground truth target sub-image; obtaining at least one second sub-image from the current frame image based on the at least one first offset position, wherein the size of the at least one second sub-image is the same as the size of the first ground truth target sub-image; and tracking the target object in the second tracking stage based on the at least one second sub-image and the first filter. If both the first and second tracking stages fail, a different filter than the first filter will be used for tracking in the next tracking stage.

2. The target tracking method according to claim 1, characterized in that, The first truth target subgraph is transformed at least once to obtain at least one transformed first truth target subgraph, including: A second truth target submap is obtained by shrinking the first truth target submap, and / or a third truth target submap is obtained by enlarging the first truth target submap.

3. The target tracking method according to claim 2, characterized in that, Also includes: A first training set is constructed based on the first ground truth target subgraph, a second training set is constructed based on the second ground truth target subgraph, and a third training set is constructed based on the third ground truth target subgraph; A first filter is obtained by training the first training set, a second filter is obtained by training the second training set, and a third filter is obtained by training the third training set.

4. The target tracking method according to claim 3, characterized in that, Also includes: Tracking failed; A third sub-image is obtained from the current frame image based on the position, wherein the third sub-image and the second ground truth target sub-image have the same size; The target object is tracked based on the third subgraph and the second filter; If tracking fails, a second offset is made to the position to obtain at least one second offset position, wherein the distance of the second offset is determined by the second ground truth target sub-map; At least one fourth sub-image is obtained from the current frame image based on the at least one second offset position, wherein the size of the at least one fourth sub-image is the same as the size of the second ground truth target sub-image; The target object is tracked based on the at least one fourth subgraph and the second filter.

5. The target tracking method according to claim 4, characterized in that, Also includes: Tracking failed; A fifth sub-image is obtained from the current frame image based on the position, wherein the fifth sub-image and the third ground truth target sub-image of the target object have the same size; The target object is tracked based on the fifth sub-graph and the third filter; If tracking fails, a third offset is performed on the position to obtain at least one third offset position, wherein the distance of the third offset is determined by the third ground truth target sub-map; At least one sixth sub-image is obtained from the current frame image based on the at least one third offset position, wherein the size of the at least one sixth sub-image is the same as the size of the third truth target sub-image; The target object is tracked based on the at least one sixth subgraph and the third filter.

6. The target tracking method according to any one of claims 1-5, characterized in that, Determining tracking failure includes: obtaining the sidelobe ratio of the corresponding subgraph based on the filter, and determining tracking failure based on the sidelobe ratio of the subgraph being less than the sidelobe ratio threshold.

7. The target tracking method according to claim 6, characterized in that, The sidelobe ratio of the corresponding subgraph obtained from the filter includes: Obtain the discrete Fourier transform result of the corresponding subgraph; Perform an inverse discrete Fourier transform on the product of the filter and the discrete Fourier transform result to obtain an intermediate image; The sidelobe ratio of the corresponding sub-image is obtained based on the peak value, mean, and standard deviation of the intermediate image.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the target tracking method according to any one of claims 1 to 7.

9. An electronic device, characterized in that, The electronic device includes: A memory that stores a computer program; The processor, which is communicatively connected to the memory, executes the target tracking method according to any one of claims 1 to 7 when it invokes the computer program.

Images