Object tracking method and device, automatic production system and computer equipment

Abstract

The invention relates to an object tracking method and device, an automatic production system, computer equipment, a computer readable storage medium and a computer program product, and the method comprises the steps: obtaining regional environment information, carrying out the colorization processing of the regional environment information, and determining a to-be-tracked target in the marked regional environment information based on reference information. And generating a feature graph system according to the to-be-tracked target, comparing the reference information with the feature graph system, determining the consistency of the to-be-tracked target and the target object, and tracking the to-be-tracked target under the condition that the consistency meets a set threshold value. Wherein the regional environment information comprises target object information and scene information of a scene where a target object is located, and the reference information comprises target object reference information. The to-be-tracked target is identified through colorization processing, so that the to-be-tracked target is distinguished from the background, the problem of tracking loss or incomplete tracking is reduced, the to-be-tracked target is further compared with the target object by generating the feature graph system, and the tracking accuracy is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of visual tracking technology, and in particular to an object tracking method, device, automated production system, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the rapid development of artificial intelligence technology, especially deep learning, visual tracking and collaborative processing have been widely used in the field of industrial automation. For example, the single object tracking (SOT) method is used to track a single workpiece in an automated assembly line factory to facilitate the subsequent grasping, adjustment or movement of the workpiece.

[0003] However, the complexity and variability of industrial environments pose severe challenges to single target tracking technology. Faced with various equipment, workpieces, and dynamically changing environments, it becomes more difficult to clearly distinguish a target. For deep learning models, similar features in the background may cause the model's attention to be distracted, thus affecting the accurate tracking of the target. The emergence of these problems not only reduces the accuracy of target tracking, but may also lead to incomplete morphology of the tracking area, that is, the tracking box may not accurately cover the entire target or may include part of the background. This situation directly weakens the accuracy and reliability of the visual system, which in turn affects subsequent automation tasks such as robot grasping, product inspection, and quality control, resulting in insufficient accuracy in artifact tracking in industrial automation. Summary of the invention

[0004] Based on this, it is necessary to provide an object tracking method, device, automated production system, computer equipment, computer-readable storage medium and computer program product that can improve tracking accuracy in response to the above technical problems.

[0005] In a first aspect, the present application provides an object tracking method, the method comprising:

[0006] Acquire regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located;

[0007] Performing colorization processing on the regional environment information, and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information;

[0008] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0009] When the consistency meets a set threshold, the target to be tracked is tracked.

[0010] In one embodiment, the colorization processing of the regional environment information and determining the target to be tracked in the marked regional environment information based on the reference information includes:

[0011] Inputting the regional environment information into a generation subnetwork, and determining target regional information within the regional environment information;

[0012] Performing color processing on the target area information to obtain marked area environment information;

[0013] The marked regional environment information and reference information are input into a discrimination subnetwork, and based on the discrimination result, a target to be tracked in the marked regional environment information is determined.

[0014] In one embodiment, determining the target to be tracked in the marked regional environment information based on the discrimination result includes:

[0015] If the determination result is true, determining that the target area information matches the target object, and using the target area information as a target to be tracked in the regional environment information;

[0016] When the determination result is false, it is determined that the target area information does not match the target object, and it is determined that the target to be tracked in the regional environment information is empty.

[0017] In one embodiment, after inputting the annotated regional environment information and the reference information into a discrimination subnetwork and determining the target to be tracked in the annotated regional environment information based on the discrimination result, the method further includes:

[0018] Marking the obtained discrimination result on the marked regional environment information to obtain an information data packet;

[0019] The reference information is updated based on the information data packet.

[0020] In one embodiment, the method further comprises:

[0021] The generative subnetwork and discriminative subnetwork are designed based on adversarial neural networks and loss functions.

[0022] In one embodiment, generating a feature map system according to the target to be tracked, comparing the reference information with the feature map system, and determining the consistency between the target to be tracked and the target object includes:

[0023] Generate a feature map system for the target to be tracked based on an attention mechanism;

[0024] The characteristic information of the reference information is compared with the characteristic map to determine the consistency between the target to be tracked and the target object.

[0025] In one embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, and generating a feature map system for the target to be tracked based on the attention mechanism includes:

[0026] Extracting features of the target to be tracked to generate channel features;

[0027] Inputting the channel feature into the channel attention mechanism and the pixel attention mechanism to obtain a first channel weight and a second channel weight;

[0028] A feature map system is obtained according to the target to be tracked, the first channel weight and the second channel weight.

[0029] In a second aspect, the present application further provides an object tracking device, the device comprising:

[0030] An input module, used to obtain regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located;

[0031] An adaptive module, used for colorizing the regional environment information and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information;

[0032] an attention module, configured to generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0033] The output module is used to track the target to be tracked when the consistency meets a set threshold.

[0034] In a third aspect, the present application also provides an automated production system, which includes a production line, a camera device and a controller, wherein the camera device is connected to the controller, and a shooting area of ​​the camera device includes the production line and objects on the production line, and is used to obtain regional environmental information and transmit it to the controller, and the controller is used to control the camera device to track objects on the production line based on the object tracking method described in the above embodiments.

[0035] In a fourth aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0036] Acquire regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located;

[0037] Performing colorization processing on the regional environment information, and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information;

[0038] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0039] When the consistency meets a set threshold, the target to be tracked is tracked.

[0040] In a fifth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0041] Acquire regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located;

[0042] Performing colorization processing on the regional environment information, and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information;

[0043] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0044] When the consistency meets a set threshold, the target to be tracked is tracked.

[0045] In a sixth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0046] Acquire regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located;

[0047] Performing colorization processing on the regional environment information, and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information;

[0048] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0049] When the consistency meets a set threshold, the target to be tracked is tracked.

[0050] The object tracking method, device, automated production system, computer equipment, computer-readable storage medium and computer program product include obtaining regional environmental information, colorizing the regional environmental information, and determining the target to be tracked in the marked regional environmental information based on the reference information. Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, determine the consistency between the target to be tracked and the target object, and track the target to be tracked when the consistency meets the set threshold. Among them, the regional environmental information includes target object information and scene information of the scene where the target object is located, and the reference information includes target object reference information. By using colorization processing to identify the target to be tracked, it can be distinguished from the background in the environment, reducing the problem of tracking loss or incomplete tracking, and further comparing the target to be tracked with the target object by generating a feature map system, which is conducive to ensuring the accuracy of tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0052] FIG1 is a diagram of an application environment of an object tracking method according to an embodiment;

[0053] FIG2 is a schematic diagram of a flow chart of an object tracking method in one embodiment;

[0054] FIG3 is a flowchart of the steps of colorizing regional environment information and determining a target to be tracked in the marked regional environment information based on reference information in one embodiment;

[0055] FIG4 is a flowchart of a step of determining a target to be tracked in the marked regional environment information based on the discrimination result in one embodiment;

[0056] FIG5 is a schematic flow chart of an object tracking method in another embodiment;

[0057] FIG6 is a schematic diagram of the structure of an adaptive module according to an embodiment;

[0058] 7 is a flowchart of steps for generating a feature map system according to a target to be tracked, comparing reference information with the feature map system, and determining the consistency between the target to be tracked and the target object in one embodiment;

[0059] FIG8 is a schematic diagram of a flow chart of steps for generating a feature map system for a target to be tracked based on an attention mechanism in one embodiment;

[0060] FIG9 is a schematic diagram of the structure of an attention module according to an embodiment;

[0061] FIG10 is a flow chart of an object tracking method in yet another embodiment;

[0062] FIG11 is a block diagram of a device for tracking an object in one embodiment;

[0063] FIG. 12 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first resistor may be referred to as a second resistor, and similarly, a second resistor may be referred to as a first resistor. Both the first resistor and the second resistor are resistors, but they are not the same resistor.

[0066] It can be understood that the “connection” in the following embodiments should be understood as “electrical connection”, “communication connection”, etc. if the connected circuits, modules, units, etc. have electrical signals or data transmission between each other.

[0067] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0069] In today's industrial production scenes, there are many devices and workpieces, and the scenes are changeable. If you need to continuously track the movement of a target in an industrial production scene, similar features such as similar workpieces and similar environments in the background may cause the model's attention to be distracted during the tracking process. The target area to be tracked may be incomplete in shape, the tracking frame may not accurately cover the target, and the tracking frame may mistakenly regard the background area as the target, resulting in reduced target tracking accuracy. In order to solve this problem, the present application provides an object tracking method that can distinguish target objects and background objects with high accuracy and achieve tracking of the target object.

[0070] The object tracking method provided in the embodiment of the present application can be applied to different application environments. For example, the object tracking method can be applied to an automated production system, as shown in FIG1, an automated production system, the automated production system includes a production line 102, a camera 104 and a controller 106, the camera 104 is connected to the controller 106, the shooting area of ​​the camera 104 (as shown by the dotted line in the figure) includes the production line 102 and objects on the production line 102 (not shown in the figure), wherein the objects on the production line 102 include the target object to be tracked. Optionally, the category of the production line 102 is not limited, for example, it can be a production line of different electrical components, then the object on the production line 102 is the corresponding electrical component, and the target object to be tracked can be one of multiple electrical components.

[0071] Optionally, one camera device 104 may also track multiple objects in the shooting area. Furthermore, when there are more than two cameras 104 and controllers 106, it is also possible to track different objects on the same production line 102, that is, by taking different objects on the production line 102 as target objects, it is possible to track multiple objects on the production line 102.

[0072] The camera device 104 shoots the shooting area, obtains regional environmental information, and transmits it to the controller 106. The controller 106 is used to control the camera device 104 to track the object on the production line 102 based on the object tracking method described in each embodiment of the present application. Exemplarily, when the shooting area of ​​the camera device 104 can change, when the target object shot by the camera device 104 moves with the production line 102, the controller 106 can control the shooting area of ​​the camera device 104 to move with the target object to ensure that the target object is always in the shooting area of ​​the camera device 104 until the tracking is completed. Optionally, the camera device 104 can be a spherical camera or a monitor, etc., and the regional environmental information obtained by the shooting can be presented in the form of an image, which can be a grayscale image, and the corresponding camera device 104 can be a grayscale camera.

[0073] In an exemplary embodiment, as shown in FIG. 2 , an object tracking method is provided, which is described by taking the method applied to the controller 106 in FIG. 1 as an example, and includes the following steps 202 to 208 .

[0074] Step 202, obtaining regional environment information.

[0075] Among them, the regional environmental information is the image information obtained by the camera device shooting the shooting area. Since the shooting area includes the production line and the objects on the production line, it also includes the target object, and the regional environmental information obtained by shooting includes the target object information. Since the image information at this time has not yet tracked the target object, the regional environmental information can be marked as including the target object information and the scene information of the scene where the target object is located, that is, the regional environmental information includes the target object information and the scene information of the scene where the target object is located. Optionally, there may be no target object on the production line, and the regional environmental information at this time may not include the target object information. This situation is considered in the subsequent method steps and will not be discussed here.

[0076] Specifically, the controller is connected to the camera device and can obtain the regional environment information captured by the camera device. Optionally, the controller can also obtain an image from the camera device, and the controller obtains the regional environment information after grayscale, sharpening or exposure processing on the image.

[0077] Step 204 , colorize the regional environment information, and determine the target to be tracked in the marked regional environment information based on the reference information.

[0078] The reference information includes target object reference information. The reference information includes target object reference information, which may be obtained by photographing an object with the same appearance, category or characteristics as the target object by a camera device, or may be obtained by summarizing historical data through big data or AI processing. The target object reference information is highly similar to the target object information photographed by the camera device, and can be considered equal.

[0079] Specifically, the controller stores reference information and relevant data of the target object. The relevant data is not necessarily stored in the form of image information obtained by camera, and is different from the reference information. The controller first preliminarily analyzes the regional environmental information based on the relevant data of the target object, and proposes the location of the target object, performs color processing on the location, and obtains the marked regional environmental information. The color processing can color the specified area, including but not limited to brightening or covering different colors for the specified area in the grayscale image, covering other colors for the specified area in the color image, or adding fluorescent highlight marks.

[0080] After obtaining the marked regional environmental information, the controller analyzes the marked regional environmental information based on the reference information to determine whether the colorized area is the location of the target object, and correspondingly determines the target to be tracked in the marked regional environmental information.

[0081] Step 206 , generating a feature map system according to the target to be tracked, comparing the reference information with the feature map system, and determining the consistency between the target to be tracked and the target object.

[0082] After obtaining the target to be tracked, it is also necessary to confirm whether the target to be tracked matches the target object. Specifically, the controller extracts the features of the target to be tracked, establishes a feature map, compares the feature map with the reference information, and carefully distinguishes the difference between the feature map of the target to be tracked and the feature information of the reference information. Since the reference information includes the target object reference information, it can be used as a standard for comparison of the target object. Therefore, when comparing the feature map with the reference information, the consistency of the target to be tracked and the target object can be determined.

[0083] Step 208: When the consistency meets the set threshold, the target to be tracked is tracked.

[0084] The controller stores a set threshold value for comparing the target to be tracked with the target object to obtain consistency. When the controller analyzes the consistency between the target to be tracked and the target object, the set threshold value is used as a judgment basis to distinguish whether the target to be tracked is the target object. When the consistency meets the set threshold value, the target to be tracked can be considered to be the target object; when the consistency does not meet the set threshold value, the target to be tracked can be considered not to be the target object.

[0085] Furthermore, when the consistency is sufficient, that is, when the consistency between the target to be tracked and the target object meets the set threshold, the target to be tracked is characterized as the target object, and the target to be tracked can be tracked. Exemplarily, when the controller determines that the target to be tracked is the target object, it can control the camera device to follow and shoot the target to be tracked, so that the shooting area of ​​the camera device moves with the movement of the target to be tracked, thereby tracking the target to be tracked.

[0086] Optionally, when the consistency is insufficient, that is, when the consistency between the target to be tracked and the target object does not meet a set threshold, it indicates that the target to be tracked is not the target object, and there is no need to track the target to be tracked, and no tracking control is required.

[0087] The object tracking method includes obtaining regional environmental information, colorizing the regional environmental information, and determining the target to be tracked in the marked regional environmental information based on the reference information. A feature map is generated according to the target to be tracked, and the reference information and the feature map are compared to determine the consistency between the target to be tracked and the target object. When the consistency meets the set threshold, the target to be tracked is tracked. The regional environmental information includes the target object information and the scene information of the scene where the target object is located, and the reference information includes the target object reference information. By using colorization processing to identify the target to be tracked, it can be distinguished from the background in the environment, reducing the problem of tracking loss or incomplete tracking, and further comparing the target to be tracked with the target object by generating a feature map, which is conducive to ensuring the accuracy of tracking.

[0088] In an exemplary embodiment, as shown in FIG. 3 , step 204 includes steps 302 to 306 .

[0089] Step 302: input the regional environment information into the generation sub-network to determine the target region information in the regional environment information.

[0090] Among them, the generation subnetwork is a part of the adaptive neural network (adaptive module), which is established based on the relevant data of the target object and builds a network environment that can distinguish regional environmental information and mark information similar to the target object from the input information.

[0091] Specifically, the regional environmental information obtained by the controller is first input into the generation sub-network of the adaptive module, and the generation sub-network analyzes the regional environmental information to obtain the target regional information. The target regional information is the regional information most similar to the location of the target object, and is also the regional information in the regional environmental information where the target object is most likely to exist.

[0092] Step 304: colorize the target area information to obtain labeled area environment information.

[0093] The adaptive module also includes a coloring subnetwork, and the coloring subnetwork and the generating subnetwork can be combined into a generator to complete the labeling of the regional environmental information. Specifically, the obtained target area information is input into the coloring subnetwork for colorization processing, and the target area information is colored, which can be highlighted or marked with a special color to distinguish the target area information from other information in the regional environmental information. After completing the processing in the generating subnetwork and the coloring subnetwork, that is, completing the processing of the generator in the adaptive module, the regional environmental information is processed into the labeled regional environmental information.

[0094] In step 306, the labeled regional environment information and the reference information are input into the discrimination subnetwork, and based on the discrimination result, the target to be tracked in the labeled regional environment information is determined.

[0095] The adaptive module also includes a discriminant subnetwork, also known as a discriminator or semantic subnetwork, which can discriminate between two types of input information to obtain a discrimination result. The discrimination process is a true or false judgment. The discriminator can mark one type of input information as "true" to discriminate another type of input information to obtain a true or false discrimination result. Specifically, the labeled regional environmental information and reference information are both input into the discriminant subnetwork, and the discriminant subnetwork marks the reference information as "true", discriminates the labeled regional environmental information, and obtains a discrimination result. After obtaining the discrimination result, the controller can determine the target to be tracked in the labeled regional environmental information based on the discrimination result.

[0096] Exemplarily, different targets to be tracked are determined based on different discrimination results. In an exemplary embodiment, as shown in FIG4 , the step 306 of determining the target to be tracked in the marked regional environment information based on the discrimination result includes steps 402 to 404 .

[0097] Step 402: if the judgment result is true, determine that the target region information matches the target object, and use the target region information as the target to be tracked in the marked regional environment information.

[0098] Specifically, when the discrimination result of the discriminant subnetwork is true, the target area information in the marked regional environment information is sufficiently similar to the target object reference information in the reference information, that is, the target area information matches the target object, then the target area information can be used as the target to be tracked in the marked regional environment information to facilitate subsequent object tracking.

[0099] Step 404: when the judgment result is false, it is determined that the target area information does not match the target object, and it is determined that the target to be tracked in the regional environment information is empty.

[0100] Specifically, when the discrimination result of the discrimination subnetwork is false, it means that the target area information in the marked regional environment information is not similar enough to the target object reference information in the reference information, that is, the target area information does not match the target object. Then, the target area information in the marked regional environment information is not the information corresponding to the target object, indicating that there is no target object in the regional environment information, and it can be determined that the target to be tracked is empty.

[0101] In this embodiment, by using the adaptive module, the regional environment information is passed through the generator's generation subnetwork and coloring subnetwork to obtain the target region information after colorization, and then the marked regional environment information is judged by the judge, and the target to be tracked is determined according to the judgment result. By coloring the corresponding information of the target object based on the adaptive module and further judging whether it is accurate, the target object and the background environment are distinguished, the difference between the target object and the background environment is increased, and the integrity of the target object is guaranteed, which is conducive to achieving object tracking and improving tracking accuracy.

[0102] In an exemplary embodiment, the labeled regional environment information can also be used for training in the adaptive module. As shown in FIG5 , after step 306 , the object tracking method further includes steps 502 to 504 .

[0103] Step 502, marking the obtained discrimination result with the marked regional environment information to obtain an information data packet.

[0104] Specifically, after the discrimination subnetwork discriminates the marked regional environment information to obtain a discrimination result, the corresponding discrimination result can be marked on the marked regional environment information, and data is packaged to obtain a message data packet.

[0105] Step 504: Update the reference information based on the information data packet.

[0106] It can be known that the labeled regional environmental information marked with the discrimination result is relatively true and reliable, so the information data packet can be used as reference information to update the reference information. After the reference information is updated, the updated reference information can be used as a reference for other labeled regional environmental information. The updated reference information is input, and the discrimination subnetwork marks the updated reference information as "true" to discriminate other labeled regional environmental information.

[0107] The above steps can be looped to discriminate multiple labeled regional environmental information, which is conducive to judging the difference between the labeled regional environmental information obtained after the generator colorization processing and the real data through the discriminant sub-network, and using the discriminant sub-network to continuously constrain the quality of the generator colorization processing until the discriminator cannot distinguish between the colorized image generated by the generator colorization processing and the real color image.

[0108] For ease of understanding, the above steps are explained by taking the adaptive module structure diagram of Figure 6 as an example. As shown in Figure 6, the random variables include different regional environmental information, wherein each regional environmental information may include target object information or may not include target object information. The regional environmental information is passed through the generator to determine the target area information and colorize the target area information to obtain the labeled regional environmental information. The real data includes reference information and information data packets. The information data packets can be empty. When the information data packets are not empty, they are regarded as information data packets to update the reference information. The labeled regional environmental information and the real data are both input into the discriminator, and the discrimination result is output as true or false. The discrimination result and the labeled regional environmental information are combined to obtain the information data packet, which is used as the real data to update the reference information and complete the training of the adaptive module.

[0109] In this embodiment, after obtaining the discrimination result, an information data packet is obtained and the reference information is updated. The reference information is used as the basis for discriminating other regional environmental information. In this way, the adaptive module is gradually trained in a cycle, which is beneficial to optimizing the accuracy of colorization processing, increasing the discrimination accuracy of the target to be tracked, and thus increasing the accuracy of object tracking.

[0110] In an exemplary embodiment, the object tracking method further includes step 602: designing a generative subnetwork and a discriminative subnetwork based on an adversarial neural network and a loss function.

[0111] Specifically, step 602 is before step 302, and can be before step 202 or after step 202. Before using the generative sub-network and the discriminative sub-network, it is necessary to first use a generative adversarial network (GAN) as a basis, establish a loss function according to the colorization processing requirements, and design the required generative sub-network and discriminative sub-network.

[0112] Furthermore, the generator and discriminator of the adaptive module can also be designed based on the adversarial neural network, and a loss function can be added to reduce the loss of the generator to a minimum and increase the loss of the discriminator to a maximum, thereby ensuring the accuracy of the target to be tracked. In combination with the adaptive module shown in Figure 6, the loss function is as follows:

[0113]

[0114] In the formula, E x~Pdata（x） [log(D(x))] represents the real data loss, E z~Pdata(x) [log(1-D(G(z)))] represents the loss of generated data, x represents the real sample, z represents the labeled sample, G and D represent the generator and discriminator respectively, P data()represents the sample distribution and E represents the expectation.

[0115] In this embodiment, by designing an adaptive module based on an adversarial neural network and adding a loss function, the losses of the generator and the discriminator are adjusted, which is beneficial to ensuring the accuracy of colorization processing on the target area information, ensuring the integrity of the colored area of ​​the target object, and improving the accuracy of object tracking.

[0116] In an exemplary embodiment, as shown in FIG. 7 , step 206 includes steps 702 to 704 .

[0117] Step 702: Generate a feature map system for the target to be tracked based on the attention mechanism.

[0118] Among them, the attention mechanism is a mechanism used in deep learning. It is similar to the attention mechanism of human vision. When observing things, people will selectively pay attention to more important information, which is called attention. By continuously paying attention to a key position to obtain more information and ignoring other useless information, this visual attention mechanism improves the efficiency and accuracy of information processing. The attention mechanism in deep learning is to focus on important points among more information, select key information, and ignore other unimportant information.

[0119] In this embodiment, the attention mechanism is part of the attention module and also part of the target tracking network. Specifically, the target to be tracked is input into the target tracking network, the attention mechanism is used to extract features of the target to be tracked, and a feature graph is generated based on the extracted features.

[0120] Step 704 , comparing the feature information of the reference information with the feature map to determine the consistency between the target to be tracked and the target object.

[0121] Specifically, the reference information contains feature information of the target object, and the reference information is also input into the target tracking network (attention module) as a reference of the feature map system, and compared with the feature map system to determine the consistency between the target to be tracked and the target object.

[0122] Furthermore, the feature information of the target object includes multiple features of the target object, and the feature map also includes multiple features of the target to be tracked. The attention module compares the features separately to obtain the consistency between the target to be tracked and the target object. If the consistency is sufficient, that is, the consistency between the target to be tracked and the target object meets the set threshold, the target to be tracked is characterized as the target object, and the target to be tracked can be tracked.

[0123] In this embodiment, the attention module is used to analyze the target to be tracked, and the consistency between the target to be tracked and the target object is further determined based on the feature map system, thereby further improving the accuracy of object tracking.

[0124] In an exemplary embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, that is, the attention module includes channel attention (CA) and pixel attention (PA). As shown in FIG8 , step 702 includes steps 802 to 806 .

[0125] Step 802: extract features of the target to be tracked and generate channel features.

[0126] Specifically, the attention module can include multiple channels, which correspond to different mechanisms, such as channel attention mechanism and pixel attention mechanism. Different channels consider the target to be tracked based on different dimensions and can adapt to different target objects. Before generating the feature map system, it is necessary to generate different channel features based on different channels, and obtain channel attention through the channel attention mechanism in order to obtain the feature map system.

[0127] Exemplarily, the average pooling method can be used to introduce global information into the channel descriptor, and the feature map size changes from C·H·W to C·1·1. The formula is as follows:

[0128]

[0129] In the formula, subscript c represents the channel, H p is the global pooling function, X c (i,j) represents the cth channel X c The value output at (i,j).

[0130] Step 804: input the channel features into the channel attention mechanism and the pixel attention mechanism to obtain the first channel weight and the second channel weight.

[0131] Specifically, the channel features are input into different channels, that is, based on different attention mechanisms, the corresponding channel weights are determined. The correspondence between the first channel weight and the second channel weight and the channel attention mechanism and the pixel attention mechanism is not fixed. Optionally, the channel features are input into the channel attention mechanism to obtain the first channel weight, and the channel features are input into the pixel attention mechanism to obtain the second channel weight. Different weights can be generated according to the importance of different channels.

[0132] Exemplarily, the attention module includes channel attention (CA) and pixel attention (PA), which are sequentially connected and processed hierarchically. As shown in FIG9 , Sigmoid and ReLU are activation functions, Conv is a convolutional layer, and the number of convolutional layers is two, and Pooling is a pooling layer.

[0133] By inputting the obtained channel features into two convolutional layers and two activation functions, the weights of different channels can be obtained. The formula is as follows:

[0134]

[0135] In the formula, σ represents the activation function Sigmoid, and δ represents the activation function ReLU.

[0136] Step 806, obtaining a feature map system according to the target to be tracked, the first channel weight and the second channel weight.

[0137] Specifically, after determining the weights of each channel, the channel attention processing can be completed. The target to be tracked is multiplied by the corresponding channel weight in order to obtain the output of the channel attention:

[0138]

[0139] At the same time, the output of channel attention is used as the input of pixel attention, input into the two convolutional layers of pixel attention for downsampling, and processed based on the same activation function as channel attention to obtain the output of pixel attention:

[0140]

[0141] Finally, the output of the channel attention is multiplied by the output of the pixel attention to obtain the feature map system of the target to be tracked.

[0142] In this embodiment, a target tracking network is established by combining channel attention and pixel attention, and the target to be tracked is further analyzed and judged. Based on the obtained feature map and reference information, judgment and analysis are performed, multiple comparisons and repeated confirmations are performed to ensure that the target to be tracked is consistent with the target object, thereby ensuring the accuracy of object tracking.

[0143] Based on the same technical concept, the present application also provides an automated production system, as shown in FIG1, an automated production system, the automated production system includes a production line 102, a camera 104 and a controller 106, the camera 104 is connected to the controller 106, the shooting area of ​​the camera 104 (as shown by the dotted line in the figure) includes the production line 102 and objects on the production line 102 (not shown in the figure), wherein the objects on the production line 102 include the target object to be tracked. Optionally, the category of the production line 102 is not limited, for example, it can be a production line of different electrical components, then the object on the production line 102 is the corresponding electrical component, and the target object to be tracked can be one of multiple electrical components.

[0144] Optionally, one camera device 104 may also track multiple objects in the shooting area. Furthermore, when there are more than two cameras 104 and controllers 106, it is also possible to track different objects on the same production line 102, that is, by taking different objects on the production line 102 as target objects, it is possible to track multiple objects on the production line 102.

[0145] The camera device 104 shoots the shooting area, obtains the regional environmental information, and transmits it to the controller 106. The controller 106 is used to control the camera device 104 to track the object on the production line 102 based on the object tracking method described in the above embodiments. Exemplarily, when the shooting area of ​​the camera device 104 can change, when the target object shot by the camera device 104 moves with the production line 102, the controller 106 can control the shooting area of ​​the camera device 104 to move with the target object to ensure that the target object is always in the shooting area of ​​the camera device 104 until the tracking is completed. Optionally, the camera device 104 can be a spherical camera or a monitor, etc., and the regional environmental information obtained by the shooting can be presented in the form of an image, which can be a grayscale image, and the corresponding camera device 104 can be a grayscale camera.

[0146] In order to better understand the above solution, in combination with the application scenario shown in FIG1 , a detailed explanation is given below in combination with a specific embodiment.

[0147] In one embodiment, the object tracking method is shown in FIG10. First, the camera device captures an image and transmits it to the controller to obtain regional environmental information, and then the regional environmental information is input into the adaptive module. When the discrimination result of the adaptive module's discriminator is false, it means that the target coloring has failed, and the target to be tracked is empty, and the object tracking ends; when the discrimination result of the adaptive module's discriminator is true, it means that the target coloring has succeeded, and the target to be tracked is the target area information in the regional environmental information. The target to be tracked is input into the attention tracking network (attention module), and the tracked target object is obtained, and compared with the previous object coloring area to determine the consistency. When consistent, the target to be tracked is confirmed to be the target object, and object tracking can be achieved; when inconsistent, return to re-input the attention tracking network and analyze again.

[0148] In this embodiment, an attention module consisting of channel attention and pixel attention is added to the target tracking network to increase the attention to the details of the target object. By combining channel attention and pixel attention, the shape, outline and color of the target object are given higher attention, which effectively solves the target tracking error phenomenon caused by processing targets with high similarity, and effectively improves the accuracy of single target tracking.

[0149] At the same time, an adaptive module is added on the basis of the above. This module, as the preprocessing stage of the algorithm, is used to colorize the target object and effectively limit the range of the target object, laying the foundation for the subsequent improvement of target tracking accuracy. At the same time, it effectively reduces the interference of complex background and better highlights the object contour, thereby avoiding the problem of accurate target tracking but incomplete target area and incomplete tracking of the object due to the influence of complex environment, laying the foundation for the accuracy of subsequent tasks such as positioning and insertion of the target object.

[0150] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0151] Based on the same inventive concept, the embodiment of the present application also provides an object tracking device for implementing the object tracking method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more object tracking device embodiments provided below can refer to the limitations of the object tracking method above, and will not be repeated here.

[0152] In an exemplary embodiment, as shown in FIG. 11 , an object tracking device is provided, including: an input module 1120, an adaptive module 1140, an attention module 1160 and an output module 1180, wherein:

[0153] The input module 1120 is used to obtain regional environment information; the regional environment information includes target object information and scene information of the scene where the target object is located;

[0154] The adaptive module 1140 is used to perform color processing on the regional environment information and determine the target to be tracked in the marked regional environment information based on the reference information; the reference information includes the target object reference information;

[0155] An attention module 1160 is used to generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0156] The output module 1180 is used to track the target to be tracked when the consistency meets the set threshold.

[0157] In one embodiment, the adaptive module 1140 is further used to input the regional environment information into the generation sub-network to determine the target regional information in the regional environment information. The target regional information is colorized to obtain the marked regional environment information, and the marked regional environment information and the reference information are input into the discrimination sub-network, and based on the discrimination result, the target to be tracked in the marked regional environment information is determined.

[0158] In one of the embodiments, the adaptive module 1140 is also used to determine that the target area information matches the target object when the judgment result is true, and use the target area information as the target to be tracked in the regional environment information; when the judgment result is false, determine that the target area information does not match the target object, and determine that the target to be tracked in the regional environment information is empty.

[0159] In one embodiment, the object tracking device also includes an adaptive update module, which is used to input the labeled regional environment information and reference information into the discrimination subnetwork in the adaptive module 1140, and after determining the target to be tracked in the labeled regional environment information based on the discrimination result, mark the obtained discrimination result on the labeled regional environment information to obtain an information data packet, and update the reference information based on the information data packet.

[0160] In one embodiment, the object tracking device also includes an adaptive generation module, which is used to design a generation subnetwork and a discrimination subnetwork based on an adversarial neural network and a loss function before the adaptive module 1140 inputs the regional environment information into the generation subnetwork, filters the target area information in the marked regional environment information, and performs colorization on the target area information to obtain the labeled regional environment information.

[0161] In one embodiment, the attention module 1160 is further configured to generate a feature map for the target to be tracked based on the attention mechanism, and compare the feature information of the reference information with the feature map to determine the consistency between the target to be tracked and the target object.

[0162] In one embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, and the attention module 1160 is further used to extract features of the target to be tracked, generate channel features, input the channel features into the channel attention mechanism and the pixel attention mechanism, and obtain a first channel weight and a second channel weight. According to the target to be tracked, the first channel weight and the second channel weight, a feature map system is obtained.

[0163] Each module in the above object tracking device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0164] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be shown in FIG12. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, an object tracking method is implemented. The display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0165] Those skilled in the art will understand that the structure shown in FIG. 12 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0166] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0167] Acquire regional environment information; the regional environment information includes target object information and scene information of the scene where the target object is located;

[0168] The regional environment information is colorized, and the target to be tracked in the marked regional environment information is determined based on the reference information; the reference information includes the target object reference information;

[0169] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0170] When the consistency meets the set threshold, the target to be tracked is tracked.

[0171] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0172] The regional environment information is input into the generation subnetwork to determine the target regional information within the regional environment information. The target regional information is colorized to obtain the labeled regional environment information, and the labeled regional environment information and reference information are input into the discrimination subnetwork. Based on the discrimination result, the target to be tracked within the labeled regional environment information is determined.

[0173] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0174] When the judgment result is true, it is determined that the target area information matches the target object, and the target area information is used as the target to be tracked in the regional environment information; when the judgment result is false, it is determined that the target area information does not match the target object, and the target to be tracked in the regional environment information is determined to be empty.

[0175] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0176] The obtained discrimination result is marked on the marked regional environment information to obtain an information data packet, and the reference information is updated based on the information data packet.

[0177] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0178] The generative subnetwork and discriminative subnetwork are designed based on adversarial neural networks and loss functions.

[0179] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0180] Generate a feature map system for the target to be tracked based on the attention mechanism. Compare the feature information of the reference information with the feature map system to determine the consistency between the target to be tracked and the target object.

[0181] In one embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, and the processor further implements the following steps when executing the computer program:

[0182] The target to be tracked is subjected to feature extraction to generate channel features, which are then input into the channel attention mechanism and the pixel attention mechanism to obtain the first channel weight and the second channel weight. A feature map system is obtained based on the target to be tracked, the first channel weight and the second channel weight.

[0183] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0184] Acquire regional environment information; the regional environment information includes target object information and scene information of the scene where the target object is located;

[0185] The regional environment information is colorized, and the target to be tracked in the marked regional environment information is determined based on the reference information; the reference information includes the target object reference information;

[0186] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0187] When the consistency meets the set threshold, the target to be tracked is tracked.

[0188] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0189] The regional environment information is input into the generation subnetwork to determine the target regional information within the regional environment information. The target regional information is colorized to obtain the labeled regional environment information, and the labeled regional environment information and reference information are input into the discrimination subnetwork. Based on the discrimination result, the target to be tracked within the labeled regional environment information is determined.

[0190] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0191] When the judgment result is true, it is determined that the target area information matches the target object, and the target area information is used as the target to be tracked in the regional environment information; when the judgment result is false, it is determined that the target area information does not match the target object, and the target to be tracked in the regional environment information is determined to be empty.

[0192] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0193] The obtained discrimination result is marked on the marked regional environment information to obtain an information data packet, and the reference information is updated based on the information data packet.

[0194] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0195] The generative subnetwork and discriminative subnetwork are designed based on adversarial neural networks and loss functions.

[0196] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0197] Generate a feature map system for the target to be tracked based on the attention mechanism. Compare the feature information of the reference information with the feature map system to determine the consistency between the target to be tracked and the target object.

[0198] In one embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, and when the computer program is executed by the processor, the following steps are also implemented:

[0199] The target to be tracked is subjected to feature extraction to generate channel features, which are then input into the channel attention mechanism and the pixel attention mechanism to obtain the first channel weight and the second channel weight. A feature map system is obtained based on the target to be tracked, the first channel weight and the second channel weight.

[0200] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0201] Acquire regional environment information; the regional environment information includes target object information and scene information of the scene where the target object is located;

[0202] The regional environment information is colorized, and the target to be tracked in the marked regional environment information is determined based on the reference information; the reference information includes the target object reference information;

[0203] Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object;

[0204] When the consistency meets the set threshold, the target to be tracked is tracked.

[0205] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0206] The regional environment information is input into the generation subnetwork to determine the target regional information within the regional environment information. The target regional information is colorized to obtain the labeled regional environment information, and the labeled regional environment information and reference information are input into the discrimination subnetwork. Based on the discrimination result, the target to be tracked within the labeled regional environment information is determined.

[0207] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0208] When the judgment result is true, it is determined that the target area information matches the target object, and the target area information is used as the target to be tracked in the regional environment information; when the judgment result is false, it is determined that the target area information does not match the target object, and the target to be tracked in the regional environment information is determined to be empty.

[0209] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0210] The obtained discrimination result is marked on the marked regional environment information to obtain an information data packet, and the reference information is updated based on the information data packet.

[0211] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0212] The generative subnetwork and discriminative subnetwork are designed based on adversarial neural networks and loss functions.

[0213] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0214] Generate a feature map system for the target to be tracked based on the attention mechanism. Compare the feature information of the reference information with the feature map system to determine the consistency between the target to be tracked and the target object.

[0215] In one embodiment, the attention mechanism includes a channel attention mechanism and a pixel attention mechanism, and when the computer program is executed by the processor, the following steps are also implemented:

[0216] The target to be tracked is subjected to feature extraction to generate channel features, which are then input into the channel attention mechanism and the pixel attention mechanism to obtain the first channel weight and the second channel weight. A feature map system is obtained based on the target to be tracked, the first channel weight and the second channel weight.

[0217] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0218] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0219] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An object tracking method, characterized in that: The method comprises: Acquire regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located; Performing colorization processing on the regional environment information, and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information; Generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object; When the consistency meets a set threshold, the target to be tracked is tracked.

2. The method according to claim 1, characterized in that The colorization processing of the regional environment information and determining the target to be tracked in the marked regional environment information based on the reference information includes: Inputting the regional environment information into a generation subnetwork, and determining target regional information within the regional environment information; Performing color processing on the target area information to obtain marked area environment information; The marked regional environment information and reference information are input into a discrimination subnetwork, and based on the discrimination result, a target to be tracked in the marked regional environment information is determined.

3. The method according to claim 2, characterized in that The step of determining the target to be tracked in the marked regional environment information based on the discrimination result includes: If the determination result is true, determining that the target area information matches the target object, and using the target area information as a target to be tracked in the regional environment information; When the determination result is false, it is determined that the target area information does not match the target object, and it is determined that the target to be tracked in the regional environment information is empty.

4. The method according to claim 2, characterized in that: After inputting the marked regional environment information and the reference information into the discrimination subnetwork and determining the target to be tracked in the marked regional environment information based on the discrimination result, the method further includes: Marking the obtained discrimination result on the marked regional environment information to obtain an information data packet; The reference information is updated based on the information data packet.

5. The method according to claim 2, characterized in that: The method further comprises: The generative subnetwork and discriminative subnetwork are designed based on adversarial neural networks and loss functions.

6. The method according to claim 1, characterized in that The generating a feature map system according to the target to be tracked, comparing the reference information with the feature map system, and determining the consistency between the target to be tracked and the target object includes: Generate a feature map system for the target to be tracked based on an attention mechanism; The characteristic information of the reference information is compared with the characteristic map to determine the consistency between the target to be tracked and the target object.

7. The method according to claim 6, characterized in that The attention mechanism includes a channel attention mechanism and a pixel attention mechanism. The generating a feature map system for the target to be tracked based on the attention mechanism includes: Extracting features of the target to be tracked to generate channel features; Inputting the channel feature into the channel attention mechanism and the pixel attention mechanism to obtain a first channel weight and a second channel weight; A feature map system is obtained according to the target to be tracked, the first channel weight and the second channel weight.

8. An object tracking device, characterized in that: The device comprises: An input module, used to obtain regional environmental information; the regional environmental information includes target object information and scene information of the scene where the target object is located; An adaptive module, used for colorizing the regional environment information and determining the target to be tracked in the marked regional environment information based on reference information; the reference information includes target object reference information; an attention module, configured to generate a feature map system according to the target to be tracked, compare the reference information with the feature map system, and determine the consistency between the target to be tracked and the target object; The output module is used to track the target to be tracked when the consistency meets a set threshold.

9. An automated production system, characterized in that: The automated production system includes a production line, a camera device and a controller. The camera device is connected to the controller. The shooting area of ​​the camera device includes the production line and objects on the production line. The camera device is used to obtain regional environmental information and transmit it to the controller. The controller is used to control the camera device to track objects on the production line based on the object tracking method described in any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images