Driverless Vehicle Identity Recognition and Target Tracking System, Method, Device and Electronic Equipment

Abstract

The present invention provides an identity recognition and target tracking system, method, device and electronic device for an autonomous vehicle, including: S101, obtaining the target detection result of the autonomous vehicle through a target detection algorithm; S102, training a tracker according to the target detection result of the autonomous vehicle; S103, determining the target tracking result of the autonomous vehicle in the current frame according to the position with the maximum probability in the confidence map; S104, performing information fusion on the target tracking result of the autonomous vehicle and the target detection result of the autonomous vehicle; S105, calculating the Euclidean distance between the first target box and the second target box; S106, setting a threshold according to the error between the Euclidean distances; S107, distinguishing whether the target obtained by the autonomous vehicle tracking network model for tracking the autonomous vehicle is the same as the target obtained by the target detection algorithm for tracking the autonomous vehicle through the threshold. If so, updating the tracker; S108, repeatedly executing steps S103 to S107 until the current image sequence is processed. To solve the problem that the target tracking method in the prior art is not suitable for long-term identification and tracking of targets.

Description

Technical Field

[0001] This document relates to the technical field of unmanned vehicle positioning, and particularly to an unmanned vehicle identification and target tracking system, method, device and electronic device. Background Art

[0002] With the development of robotics and deep learning technologies, object detection and object tracking technologies play an increasingly important role in military and civilian fields, and the requirements for object detection and object tracking have also become higher. They are widely used in battlefield reconnaissance, low-altitude defense, traffic monitoring, and homeland security. In complex and changeable actual combat scenarios, the type of unmanned vehicle can be identified by vision and a specified type of unmanned vehicle can be continuously tracked for a long time. Due to factors such as changes in illumination in the environment, camera jitter, target occlusion, non-linear deformation of the target, and noise interference in the background, it poses great challenges to identify and track specific unmanned vehicle targets.

[0003] At the same time, due to situations such as fast movement and motion blur, target appearance deformation, background similarity interference, illumination change, in-plane and out-of-plane rotation, scale change, occlusion, and being out of the field of view, most of the existing object tracking methods are suitable for tracking targets in a short time, and there is little research on methods for stably identifying and tracking targets for a long time. Summary of the Invention

[0004] The purpose of the present invention is to provide an unmanned vehicle identification and target tracking system, method, device and electronic device. The unmanned vehicle identification and target tracking method can solve the problem that the existing object tracking methods are not suitable for identifying and tracking targets for a long time.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] An unmanned vehicle identification and target tracking method, the method specifically includes:

[0007] S101, read the unmanned vehicle image sequence, preprocess the image sequence, and obtain the unmanned vehicle target detection result through an object detection algorithm;

[0008] S102, train a tracker according to the unmanned vehicle target detection result;

[0009] S103, perform deep feature extraction on the unmanned vehicle to be tracked through the tracker to obtain the confidence map of the image candidate region; determine the unmanned vehicle target tracking result in the current frame according to the position with the maximum probability in the confidence map;

[0010] S104, perform information fusion on the unmanned vehicle target tracking result and the unmanned vehicle target detection result;

[0011] S105. Obtain the first target box according to the unmanned vehicle tracking network model, obtain the second target box according to the target detection algorithm, and calculate the Euclidean distance between the first target box and the second target box;

[0012] S106. Set a threshold according to the error between the Euclidean distances;

[0013] S107. Determine whether the target of tracking the unmanned vehicle obtained by the unmanned vehicle tracking network model is the same as the target of tracking the unmanned vehicle obtained by the target detection algorithm through the threshold. If so, update the tracker;

[0014] S108. Loop and execute steps S103 to S107 until the current image sequence is processed.

[0015] Based on the above technical solutions, the present invention can also be improved as follows:

[0016] Further, the S101 specifically includes:

[0017] S1011. Obtain image data information through the image acquisition module;

[0018] S1012. Read the image sequence according to the image number information and preprocess the image sequence;

[0019] S1013. Perform target detection on the unmanned vehicle to be tracked through the target detection algorithm, and determine whether there is a target vehicle to be tracked. If so, obtain the target detection result of the unmanned vehicle, and the target detection result of the unmanned vehicle includes the category information and bounding box information of the target.

[0020] Further, the S102 specifically includes:

[0021] S1021. Obtain the position information of the unmanned vehicle to be tracked according to the category information and bounding box information;

[0022] S1022. Make a dataset of the pictures of the unmanned vehicle to be tracked through an online dataset annotation website;

[0023] S1023. Initialize the recurrent neural network and the siamese neural network according to the target detection result of the unmanned vehicle;

[0024] S1024. Integrate the recurrent neural network and the siamese neural network to construct a tracker.

[0025] Further, the S103 specifically includes:

[0026] S1031. Introduce the temporal and spatial attention mechanism into the recurrent neural network, extract the depth feature information, and establish the motion model and the observation model;

[0027] S1032. Generate a set of candidate regions according to the motion model;

[0028] S1033. Determine the target tracking result of the driverless vehicle in the current frame according to the observation model.

[0029] A driverless vehicle identification and target tracking system includes:

[0030] A detection model, which is used to read the image sequence of the driverless vehicle, preprocess the image sequence, and obtain the target detection result of the driverless vehicle through a target detection algorithm;

[0031] A tracker, which is connected to the driverless vehicle detection model, and is used to extract deep features of the driverless vehicle to be tracked to obtain a confidence map of the image candidate region; determine the target tracking result of the driverless vehicle in the current frame according to the position with the maximum probability of the confidence map;

[0032] A processing module, which is connected to the tracker and the detection model, and is used to fuse the information of the target tracking result of the driverless vehicle and the target detection result of the driverless vehicle, obtain a first target box according to the driverless vehicle tracking network model, obtain a second target box according to the target detection algorithm, and calculate the Euclidean distance between the first target box and the second target box;

[0033] A control module, which is connected to the processing module, and is used to set a threshold according to the error between the Euclidean distances, and distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model and the target of tracking the driverless vehicle obtained by the target detection algorithm are the same through the threshold. If so, update the tracker.

[0034] Further, the detection module is further used for:

[0035] Read the image sequence according to the image number information, and preprocess the image sequence;

[0036] Perform target detection on the driverless vehicle to be tracked through a target detection algorithm, and judge whether there is a target vehicle to be tracked. If so, obtain the target detection result of the driverless vehicle, and the target detection result of the driverless vehicle includes the category information and bounding box information of the target.

[0037] Further, the tracker is further used for:

[0038] Obtain the position information of the driverless vehicle to be tracked according to the category information and the bounding box information;

[0039] Make a dataset of the pictures of the driverless vehicle to be tracked through an online dataset annotation website;

[0040] Initialize the recurrent neural network and the siamese neural network according to the target detection result of the driverless vehicle;

[0041] Construct a tracker by fusing a recurrent neural network and a siamese neural network.

[0042] Furthermore, the unmanned vehicle identification and target tracking system further includes a motion model and an observation model, and the motion model and the observation model are connected to the tracker;

[0043] The motion model is used to generate a set of candidate regions, and the observation model is used to determine the unmanned vehicle target tracking result in the current frame.

[0044] An unmanned vehicle identification and target tracking device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the unmanned vehicle identification and target tracking method are implemented.

[0045] An electronic device stores an implementation program for information transmission. When the program is executed by a processor, the steps of the unmanned vehicle identification and target tracking method are implemented.

[0046] The present invention has the following advantages:

[0047] The unmanned vehicle identification and target tracking method of the present invention uses a tracking algorithm after fusing a recurrent neural network (RNN) and a siamese neural network, introduces a temporal and spatial attention mechanism, and combines a target detection algorithm to achieve long-term stable tracking of an unmanned vehicle. The method of fusing information between the tracker and the detector further improves the accuracy of the tracked target by fusing the target information obtained by the tracker and the detector. In practical application scenarios, the tracker can be used alone to directly track the unmanned vehicle, or the target information of target detection can be fused to achieve unmanned vehicle tracking. Different methods can be combined to meet different task requirements. The target detection algorithm is used to solve the re-identification problem. When the target of the unmanned vehicle to be tracked is lost due to occlusion or being out of the visual range, etc., the target detection algorithm is used to detect all videos, and after finding the corresponding target category, tracking is performed to achieve the task of autonomously retrieving the unmanned vehicle to be tracked and quickly tracking the specified unmanned vehicle, while greatly improving the generalization ability of the unmanned vehicle detection and tracking model in different scenarios and reducing the misjudgment rate. It solves the problem that the target tracking method in the prior art is not suitable for long-term identification and tracking of targets. Description of the Drawings

[0048] To more clearly illustrate the technical solutions in one or more embodiments of this specification or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a flowchart of the method for unmanned vehicle identity recognition and target tracking in an embodiment of the present invention;

[0050] Figure 2 It is a flowchart of S101 in an embodiment of the present invention;

[0051] Figure 3 It is a flowchart of S102 in an embodiment of the present invention;

[0052] Figure 4 It is a flowchart of S103 in an embodiment of the present invention;

[0053] Figure 5 It is a flowchart of unmanned vehicle identity recognition and tracking in an embodiment of the present invention;

[0054] Figure 6 It is a schematic diagram of the tracking network model in an embodiment of the present invention. Detailed implementation manners

[0055] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will clearly and completely describe the technical solutions in one or more embodiments of this specification in conjunction with the accompanying drawings in one or more embodiments of this specification. Obviously, the described embodiments are only some embodiments of this specification, rather than all embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this document.

[0056] As Figure 1 shown, a method for unmanned vehicle identity recognition and target tracking, the method specifically includes:

[0057] S101, obtaining the unmanned vehicle target detection result;

[0058] In this step, the unmanned vehicle image sequence is read, the image sequence is preprocessed, and the unmanned vehicle target detection result is obtained through the target detection algorithm;

[0059] Read the first frame image of the unmanned vehicle video or image sequence, preprocess the first frame image, and quickly determine the position and category of the tracking target by using the target detection algorithm;

[0060] S102, Train the tracker;

[0061] In this step, train the tracker according to the target detection results of the driverless vehicle; use the tracking target of the current frame as the input data for tracking and put it into the tracking framework to train the tracker;

[0062] S103, Determine the target tracking result of the driverless vehicle in the current frame;

[0063] In this step, perform deep feature extraction on the driverless vehicle to be tracked through the tracker to obtain the confidence map of the image candidate region; determine the target tracking result of the driverless vehicle in the current frame according to the position with the maximum probability in the confidence map; read the next frame of image, use the trained convolutional deep feature information of the candidate region of the driverless vehicle tracker to obtain the confidence map of the image candidate region, and determine the position and confidence probability size of the driverless vehicle to be tracked in the current frame according to the position with the maximum probability in the confidence map;

[0064] S104, Information fusion;

[0065] In this step, perform information fusion on the target tracking result of the driverless vehicle and the target detection result of the driverless vehicle;

[0066] S105, Calculate the Euclidean distance;

[0067] In this step, obtain the first target box P1 according to the driverless vehicle tracking network model, obtain the second target box P2 according to the target detection algorithm, and calculate the Euclidean distance UC between the first target box and the second target box dis ;

[0068] S106, Set the threshold;

[0069] In this step, set the threshold Thread according to the error between the Euclidean distances; use it to distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model and the target of tracking the driverless vehicle obtained by the target detection algorithm are the same target. Use it to distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model and the target of tracking the driverless vehicle obtained by the target detection algorithm are the same target. The lost tracking target is retrieved again. When the tracked driverless vehicle is lost due to long-term occlusion or out of the field of view, etc., the target features can be extracted in the current frame to detect the target to be tracked again;

[0070] S107, Update the tracker;

[0071] In this step, distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model and the target of tracking the driverless vehicle obtained by the target detection algorithm are the same through the threshold. If so, update the tracker;

[0072] S108, loop and execute steps S103 - S107;

[0073] In this step, loop and execute steps S103 - S107 until the current image sequence is processed.

[0074] The unmanned vehicle identity recognition and target tracking method uses a target detection module to accurately identify the identity of the unmanned vehicle in a video or image sequence, quickly locate the state of the unmanned vehicle to be tracked, combines a target tracking module to establish a motion model and an observation model, accurately estimates the state of the current unmanned vehicle to be tracked, and realizes the accurate identification and long-term stable tracking of the identity of a specific unmanned vehicle.

[0075] As Figure 5 shown, first preprocess the input video sequence, use a target detection algorithm to quickly and accurately identify the unmanned vehicle to be tracked, obtain the category and bounding box of the unmanned vehicle to be tracked, so as to achieve the purpose of autonomously locating and tracking the position of the target. Determine whether it is the first frame currently. If it is the first frame, the tracker needs to be initialized; if it is not the first frame, skip the target tracker initialization step, directly establish a motion model for tracking the unmanned vehicle, generate a set of candidate regions that may be the target, and extract deep feature information of the unmanned vehicle to be tracked.

[0076] Establish an unmanned vehicle observation model, estimate the state of the unmanned vehicle to be tracked in the current frame, and at the same time fuse the target state information detected in real time by the target detection algorithm to accurately estimate the state of the target to be tracked in the current frame. Finally, use the current state of the target to update the unmanned vehicle observation model again and continue to track the next frame.

[0077] In a video or image sequence, it can accurately identify the identity of the unmanned vehicle, quickly locate the state of the unmanned vehicle to be tracked, and realize the accurate identification and long-term stable tracking of the identity of a specific unmanned vehicle.

[0078] Fuse the recurrent neural network (RNN) and the siamese neural network to construct an unmanned vehicle identity recognition and target tracking framework. At the same time, introduce a temporal and spatial attention mechanism into the recurrent neural network (RNN) to extract deep feature information, thereby establishing a motion model and an observation model to estimate the state of the current unmanned vehicle to be tracked.

[0079] Solve the re-identification problem. When the target of the unmanned vehicle to be tracked is lost due to occlusion or out of the visual range, etc., then detect all videos, find the corresponding target category and then track it, realize the task of autonomously retrieving the unmanned vehicle to be tracked and quickly tracking the specified unmanned vehicle, and at the same time greatly improve the generalization ability of the unmanned vehicle detection and tracking model in different scenarios and reduce the misjudgment rate.

[0080] Based on the above technical solution, the present invention can also be improved as follows:

[0081] As Figure 2 shown, further, the S101 specifically includes:

[0082] S1011, obtaining image data information;

[0083] In this step, the image data information is obtained through an image acquisition module;

[0084] S1012, preprocessing the image sequence;

[0085] In this step, the image sequence is read according to the image data information, and the image sequence is preprocessed;

[0086] S1013, obtaining the target detection result of the driverless vehicle;

[0087] In this step, the target detection algorithm is used to perform target detection on the driverless vehicle to be tracked, and it is judged whether there is a target vehicle to be tracked. If so, the target detection result of the driverless vehicle is obtained, and the target detection result of the driverless vehicle includes the category information and bounding box information of the target.

[0088] As Figure 3 shown, the S102 specifically includes:

[0089] S1021, obtaining the position information of the driverless vehicle to be tracked;

[0090] In this step, the position information of the driverless vehicle to be tracked is obtained according to the category information and the bounding box information; the position [x lp , y lp , x rd , y rd of the driverless vehicle to be tracked is determined, replacing the manual method of drawing a bounding box with the mouse to select the position of the target;

[0091] S1022, making a data set of the pictures of the driverless vehicle to be tracked;

[0092] In this step, a data set of the pictures of the driverless vehicle to be tracked is made through an online data set annotation website;

[0093] S1023, initializing the recurrent neural network and the siamese neural network;

[0094] In this step, the recurrent neural network and the siamese neural network are initialized according to the target detection result of the driverless vehicle;

[0095] S1024, constructing a tracker.

[0096] As Figure 6 shown, the tracking algorithm after fusing the recurrent neural network RNN and the siamese neural network Siamese is constructed as follows:

[0097] Create a dataset of images of the unmanned vehicle to be tracked through the MKAESENSE online dataset annotation website, and use the object detection algorithm to extract the depth features of the unmanned vehicle to be tracked, obtaining an unmanned vehicle detection model.

[0098] When reading in a video or image sequence in real time, detect the category and location information of the unmanned vehicle to be tracked through the model, and take the category of the unmanned vehicle and the target bounding box as the output of the detection result. Provide input for the initialization of the tracker after the fusion of the recurrent neural network RNN and the siamese neural network Siamese.

[0099] Since the target template of a single Siamese neural network is fixed, it is prone to error accumulation. When the appearance of the target changes drastically, it is difficult for the model to adapt and it is impossible to continue to improve its robustness. Therefore, fuse the recurrent neural network RNN and the siamese neural network Siamese to construct a new tracking model, as Figure 6 shown, introducing an attention mechanism and a recurrent neural network.

[0100] Most of the object tracking methods based on convolutional neural networks usually ignore the temporal sequence of the video itself and the context information of the surrounding area of the target, and only focus on the appearance and semantic features of the target to be tracked and the area information near the target to be tracked for modeling. In order to enhance the modeling of temporal continuity and spatial information. Therefore, use the recurrent neural network RNN to model on the image temporal sequence and spatial tasks.

[0101] The purpose of introducing the attention mechanism is to enable the tracker to ignore irrelevant information and focus on key information, and quickly obtain the most effective information. A set of weight coefficients learned autonomously in the recurrent neural network RNN is used, and a dynamic weighting method is adopted to suppress the mechanism irrelevant to the background area, enhancing the region of interest in the set image temporal sequence. Using the temporal and spatial attention mechanisms can quickly highlight the target and reduce the interference of useless background information.

[0102] In this step, fuse the recurrent neural network and the siamese neural network to construct a tracker.

[0103] As Figure 4 shown, the S103 specifically includes:

[0104] S1031, establish a motion model and an observation model;

[0105] In this step, introduce the temporal and spatial attention mechanisms in the recurrent neural network, extract the depth feature information, and establish a motion model and an observation model;

[0106] S1032, generate a set of candidate regions;

[0107] In this step, a set of candidate regions is generated according to the motion model;

[0108] S1033, determine the target tracking result of the driverless vehicle in the current frame.

[0109] In this step, according to the observation model, determine the target tracking result of the driverless vehicle in the current frame.

[0110] A driverless vehicle identification and target tracking system, comprising:

[0111] A detection model, which is used to read the image sequence of the driverless vehicle, preprocess the image sequence, and obtain the driverless vehicle target detection result through the target detection algorithm;

[0112] A tracker, which is connected to the driverless vehicle detection model, and is used to extract the deep features of the driverless vehicle to be tracked to obtain the confidence map of the image candidate region; determine the target tracking result of the driverless vehicle in the current frame according to the position with the highest probability in the confidence map;

[0113] A processing module, which is connected to the tracker and the detection model, and is used to perform information fusion on the driverless vehicle target tracking result and the driverless vehicle target detection result, obtain the first target box according to the driverless vehicle tracking network model, obtain the second target box according to the target detection algorithm, and calculate the Euclidean distance between the first target box and the second target box;

[0114] A control module, which is connected to the processing module, and is used to set a threshold according to the error between the Euclidean distances, and distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model is the same as the target of tracking the driverless vehicle obtained by the target detection algorithm through the threshold. If so, update the tracker.

[0115] Furthermore, the detection module is further used for:

[0116] Read the image sequence according to the image number information, and preprocess the image sequence;

[0117] Perform target detection on the driverless vehicle to be tracked through the target detection algorithm, and determine whether there is a target vehicle to be tracked. If so, obtain the driverless vehicle target detection result, and the driverless vehicle target detection result includes the category information and bounding box information of the target.

[0118] Furthermore, the tracker is further used for:

[0119] Obtain the position information of the driverless vehicle to be tracked according to the category information and the bounding box information;

[0120] Make a dataset of the pictures of the driverless vehicle to be tracked through an online dataset annotation website;

[0121] The recurrent neural network and the siamese neural network are initialized according to the object detection results of the driverless vehicle;

[0122] The recurrent neural network and the siamese neural network are fused to construct a tracker.

[0123] Furthermore, the driverless vehicle identity recognition and object tracking system further includes a motion model and an observation model, and the motion model and the observation model are connected to the tracker;

[0124] The motion model is used to generate a set of candidate regions, and the observation model is used to determine the object tracking result of the driverless vehicle in the current frame.

[0125] A driverless vehicle identity recognition and object tracking device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the driverless vehicle identity recognition and object tracking method are implemented.

[0126] An electronic device stores an implementation program for information transmission thereon. When the program is executed by a processor, the steps of the driverless vehicle identity recognition and object tracking method are implemented.

[0127] The process of using the robot dynamic obstacle removal system is as follows:

[0128] When in use, read the image sequence of the driverless vehicle, preprocess the image sequence, and obtain the object detection result of the driverless vehicle through an object detection algorithm; train a tracker according to the object detection result of the driverless vehicle; perform deep feature extraction on the driverless vehicle to be tracked through the tracker to obtain a confidence map of the image candidate region; determine the object tracking result of the driverless vehicle in the current frame according to the position with the maximum probability of the confidence map; perform information fusion on the object tracking result of the driverless vehicle and the object detection result of the driverless vehicle; obtain a first target box according to the driverless vehicle tracking network model, obtain a second target box according to the object detection algorithm, and calculate the Euclidean distance between the first target box and the second target box; set a threshold according to the error between the Euclidean distances; distinguish whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model and the target of tracking the driverless vehicle obtained by the object detection algorithm are the same through the threshold. If so, update the tracker; loop until the current image sequence is processed.

[0129] It should be noted that the embodiment of the storage medium in this specification and the embodiment of the blockchain-based service providing method in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the corresponding blockchain-based service providing method described above, and the repeated parts will not be elaborated.

[0130] The above description has been directed to specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired result. In certain implementations, multitasking and parallel processing are also possible or may be advantageous.

[0131] In the 1930s, it was quite obvious to distinguish whether an improvement in a technology was an improvement in hardware (e.g., improvement in circuit structures such as diodes, transistors, switches, etc.) or an improvement in software (improvement in method flows). However, with the development of technology, many improvements in method flows today can be regarded as direct improvements in hardware circuit structures. Almost all designers obtain the corresponding hardware circuit structures by programming the improved method flows into the hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented with a hardware entity module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logical function is determined by the user programming the device. Designers can program by themselves to "integrate" a digital system on a single PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called a Hardware Description Language (HDL). There is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones currently are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be clear that by simply performing a little logical programming on the method flow with the above-mentioned several hardware description languages and programming it into the integrated circuit, it is easy to obtain the hardware circuit that implements the logical method flow.

[0132] The controller can be implemented in any suitable manner. For example, the controller can take the form of, for example, a microprocessor 202 or a processor 202 and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor 202, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of the controller include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that, in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to logically program the method steps to enable the controller to be implemented in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same functions. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.

[0133] The systems, devices, modules, or units illustrated in the above embodiments can be specifically implemented by a computer chip or an entity, or by a product with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0134] For the convenience of description, when describing the above devices, they are described separately as various units according to their functions. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0135] Those skilled in the art should understand that one or more embodiments of this specification can be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0136] This specification is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor 202 of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor 202 of the computer or other programmable data processing device generate means for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0137] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0139] In a typical configuration, a computing device includes one or more processors 202 (CPUs), an input / output interface, a network interface, and memory.

[0140] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0141] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0142] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0143] One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0144] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0145] The above are only examples of this document and are not intended to limit this document. For those skilled in the art, this document may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this document shall be included within the scope of the claims of this document.

Claims

1. An identity recognition and target tracking method for driverless vehicles, characterized in that, The method specifically includes: S101. Read the image sequence of the driverless vehicle, preprocess the image sequence, and obtain the target detection result of the driverless vehicle through the target detection algorithm; S102. Train a tracker according to the target detection result of the driverless vehicle; S103. Extract deep features of the driverless vehicle to be tracked through the tracker to obtain the confidence map of the image candidate region; determine the target tracking result of the driverless vehicle in the current frame according to the position with the maximum probability of the confidence map; S104. Perform information fusion on the target tracking result of the driverless vehicle and the target detection result of the driverless vehicle; S105. Obtain the first target box according to the driverless vehicle tracking network model, obtain the second target box according to the target detection algorithm, and calculate the Euclidean distance between the first target box and the second target box; S106. Set a threshold according to the error between the Euclidean distances; S107. Determine whether the target of tracking the driverless vehicle obtained by the driverless vehicle tracking network model is the same as the target of tracking the driverless vehicle obtained by the target detection algorithm through the threshold. If so, update the tracker; S108. Loop and execute steps S103 to S107 until the current image sequence is processed, where the specific content of S101 includes: S1011. Obtain image data information through the image acquisition module; S1012. Read the image sequence according to the image data information and preprocess the image sequence; S1013. Perform target detection on the driverless vehicle to be tracked through the target detection algorithm, and determine whether there is a target vehicle to be tracked. If so, obtain the target detection result of the driverless vehicle. The target detection result of the driverless vehicle includes the category information and bounding box information of the target; the specific content of S102 includes: S1021. Obtain the position information of the driverless vehicle to be tracked according to the category information and the bounding box information; S1022. Make a dataset of the pictures of the driverless vehicle to be tracked through the online dataset annotation website; S1023. Initialize the recurrent neural network and the siamese neural network according to the target detection result of the driverless vehicle; S1024. Integrate the recurrent neural network and the siamese neural network to construct a tracker.

2. The method for unmanned vehicle identity recognition and target tracking according to claim 1, wherein the specific content of S103 includes: S1031. Introduce the temporal and spatial attention mechanism into the recurrent neural network, extract deep feature information, and establish a motion model and an observation model; S1032. Generate a set of candidate regions according to the motion model; S1033. Determine the target tracking result of the driverless vehicle in the current frame according to the observation model.

3. An unmanned vehicle identity recognition and target tracking system, characterized in that, It includes: A detection model, which is used to read the image sequence of the driverless vehicle, preprocess the image sequence, and obtain the target detection result of the driverless vehicle through the target detection algorithm; A tracker, which is connected to the detection model and is used to extract deep features of the driverless vehicle to be tracked to obtain the confidence map of the image candidate region; determine the target tracking result of the driverless vehicle in the current frame according to the position with the maximum probability of the confidence map; A processing module, which is connected to the tracker and the detection model, is configured to perform information fusion on the unmanned vehicle target tracking result and the unmanned vehicle target detection result, obtain a first target box according to the unmanned vehicle tracking network model, obtain a second target box according to the target detection algorithm, and calculate the Euclidean distance between the first target box and the second target box; A control module, which is connected to the processing module, is configured to set a threshold according to the error between the Euclidean distances, and distinguish whether the target of tracking the unmanned vehicle obtained by the unmanned vehicle tracking network model and the target of tracking the unmanned vehicle obtained by the target detection algorithm are the same through the threshold. If so, update the tracker; wherein, The detection module is further configured to: Read an image sequence according to the image data information and preprocess the image sequence; Perform target detection on the unmanned vehicle to be tracked through the target detection algorithm, and determine whether there is a target vehicle to be tracked. If so, obtain the unmanned vehicle target detection result, and the unmanned vehicle target detection result includes the category information and bounding box information of the target; The tracker is further configured to: Obtain the position information of the unmanned vehicle to be tracked according to the category information and the bounding box information; Make a dataset of the pictures of the unmanned vehicle to be tracked through an online dataset annotation website; The recurrent neural network and the siamese neural network are initialized according to the unmanned vehicle target detection result; Fuse the recurrent neural network and the siamese neural network to construct a tracker.

4. The unmanned vehicle identity recognition and target tracking system according to claim 3, characterized in that, The unmanned vehicle identity recognition and target tracking system further includes a motion model and an observation model, and the motion model and the observation model are connected to the tracker; The motion model is configured to generate a set of candidate regions, and the observation model is configured to determine the unmanned vehicle target tracking result in the current frame.

5. An identity recognition and target tracking device for driverless vehicles, characterized in that, Including: A memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the unmanned vehicle identity recognition and target tracking method according to any one of claims 1 to 2 are implemented.

6. An electronic device, characterized in that, An information transmission implementation program is stored on the electronic device. When the program is executed by the processor, the steps of the unmanned vehicle identity recognition and target tracking method according to any one of claims 1 to 2 are implemented.

Images