CN117649582B - Single-stream single-stage network target tracking method and system based on cascaded attention - Google Patents

Abstract

The present invention proposes a single-stream single-stage network target tracking method and system based on cascaded attention, the method comprising: firstly forming a single-stream single-stage overall model, inputting a template image and a search image into the single-stream single-stage overall model, performing feature extraction to obtain local feature information, and using cascaded attention to aggregate local semantic information to achieve feature enhancement, and then performing cross-attention calculation to achieve communication, obtaining a result feature map, and repeatedly extracting the result feature map several times in an iterative manner to obtain a final result feature map to predict the target location and target state, and achieve target tracking according to the target location, and at the same time determine whether to use the predicted target state as an online template in the next stage of online tracking process according to the target state. The present invention can reduce the computational redundancy of multi-head attention while coping with changes in the appearance of objects in complex scenes, thereby improving the performance of target tracking.

Description

Single-stream single-stage network target tracking method and system based on cascaded attention

Technical Field

The present invention relates to the technical field of computer vision and image processing, and in particular to a single-stream single-stage network target tracking method and system based on cascaded attention.

Background technique

Visual tracking is a fundamental research task in computer vision, which focuses on accurately locating an arbitrary target in each video frame using only its initial appearance as a reference. It is applied in various fields, including visual localization, autonomous driving systems, and smart city technologies. However, designing a robust tracker remains a major challenge due to many challenging factors in real-world scenes, such as partial occlusions, objects leaving the field of view, background clutter, viewpoint changes, and scale changes.

Currently, tracking models usually adopt a two-stream two-stage model architecture. In this approach, features from the template and the search region are extracted separately. However, this approach has certain disadvantages, mainly attributed to the high computational complexity of the traditional attention mechanism. In addition, local feature information is often ignored when extracting features with global context. Recently, single-stream architectures have emerged as a viable alternative. These architectures bring faster processing and enhanced feature fusion capabilities, achieving significant success in tracking performance. The reason behind their effectiveness lies in the model architecture's ability to establish an unhindered information flow between the template and the search region at an early stage, especially from the original image pair. This helps in extracting target-specific features and prevents the loss of discriminative information.

Transformer first proposed an encoder-decoder module based on self-attention mechanism for natural language processing. It explores long-range dependencies in sequences by calculating the attention weights of triplets. Based on its excellent feature fusion ability, the Transformer structure has been successfully applied to visual tracking and achieved encouraging results. In the Transformer-based tracker, the global context information is fully explored, however, the local information is not fully utilized. In order to improve the attention mechanism, a new attention module called cascade attention is proposed. The core idea is to enhance the diversity of features of the input attention head.

Summary of the invention

In view of the above situation, the main purpose of the present invention is to propose a single-stream single-stage network target tracking method and system based on cascaded attention to solve the above technical problems.

The present invention proposes a single-stream single-stage network target tracking method based on cascaded attention, the method comprising the following steps:

Step 1: In a single-stream single-stage framework, based on the Transformer network model and the feature enhancement module, a backbone feature extraction and fusion module is constructed. The backbone feature extraction and fusion module, the head corner module, and the fractional head prediction module constitute a single-stream single-stage overall model;

Step 2: Obtain a template image and a search image, wherein the template image includes an initial template containing a number of target objects to be tracked and a number of online templates containing target states;

Step 3: Input the template image and the search image into the single-stream single-stage overall model, and extract the local feature information corresponding to the template image and the search image through the backbone feature extraction and fusion module;

The local feature information is input into the feature enhancement module, and the local semantic information is aggregated using cascaded attention to achieve feature enhancement, and the global context information of the template image and the search image is obtained;

Perform cross-attention calculation on the global context information of the template image and the search image to achieve communication and obtain the resulting feature map;

Step 4: Segment the result feature map into a template image and a search image, and use them as the input for the next stage. Repeat step 3 several times in an iterative manner to obtain the final result feature map.

Step 5: Input the final result feature map into the head corner point module to predict the confidence score of each target position, and determine the position of the tracking target according to the confidence score to achieve target tracking;

The resulting feature map is input into the score head prediction module to predict the confidence score of each target state, and the confidence score of the target state is used to determine whether to use the predicted target state as an online template in the next stage of online tracking.

Step 6: Repeat steps 2 to 4 based on a large-scale dataset to pre-train the single-stream single-stage overall model to optimize model parameters.

Step 7: Use the trained single-stream single-stage overall model to track the target online in the video sequence.

The present invention also proposes a single-stream single-stage network target tracking system based on cascaded attention, wherein the system applies the single-stream single-stage network target tracking method based on cascaded attention as described above, and the system comprises:

Building blocks for:

In the single-stream single-stage framework, based on the Transformer network model and the feature enhancement module, a backbone feature extraction and fusion module is constructed. The backbone feature extraction and fusion module, the head corner module, and the fractional head prediction module constitute the single-stream single-stage overall model.

Learning modules for:

Acquire a template image and a search image, wherein the template image includes an initial template containing a number of desired tracking targets and a number of online templates containing target states;

The template image and the search image are input into the single-stream single-stage overall model, and the local feature information corresponding to the template image and the search image is extracted through the backbone feature extraction and fusion module;

The local feature information is input into the feature enhancement module, and the local semantic information is aggregated using cascaded attention to achieve feature enhancement, and the global context information of the template image and the search image is obtained;

Perform cross-attention calculation on the global context information of the template image and the search image to achieve communication and obtain the resulting feature map;

Extraction module for:

The resulting feature map is divided into a template image and a search image, and used as the input for the next stage. The feature extraction is repeated several times in an iterative manner to obtain the final result feature map.

Compute module for:

The final result feature map is input into the head corner module to predict the confidence score of each target position, and the position of the tracking target is determined according to the confidence score to achieve target tracking;

The resulting feature map is input into the score head prediction module to predict the confidence score of each target state, and the confidence score of the target state is used to determine whether to use the predicted target state as an online template in the next stage of online tracking.

Pre-trained modules for:

Pre-train the single-stream single-stage overall model based on large-scale datasets to optimize model parameters;

Tracking module for:

The trained single-stream single-stage holistic model is used to perform online object tracking in video sequences.

Compared with the prior art, the beneficial effects of the present invention are as follows:

1. The present invention uses cascaded attention to provide different input splits to each head, and then cascades the output features to these heads. This method not only reduces the computational redundancy of multi-head attention, but also enhances the capacity of the model by increasing the network depth.

2. The present invention introduces a score head module for online template updating, which corrects the online template image based on the predicted score of the search image. It can cope with changes in object appearance in complex scenes, enable it to better handle difficulties such as severe occlusion, scale changes and complex backgrounds in the tracking process, effectively capture temporal information and handle changes in object appearance, thereby improving target tracking performance.

Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description or learned through embodiments of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a single-stream single-stage network target tracking method based on cascaded attention proposed by the present invention;

FIG2 is a structural diagram of a single-stream single-stage network target tracking framework based on a cascaded attention module proposed by the present invention;

FIG3 is a schematic diagram of a feature enhancement module of the present invention;

FIG4 is a schematic diagram of the cascaded attention of the present invention;

FIG5 is a schematic diagram of the structure of a single-stream single-stage network target tracking system based on cascaded attention proposed by the present invention.

Detailed ways

Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

These and other aspects of the embodiments of the present invention will be apparent with reference to the following description and accompanying drawings. In these descriptions and accompanying drawings, some specific implementations of the embodiments of the present invention are specifically disclosed to represent some ways of implementing the principles of the embodiments of the present invention, but it should be understood that the scope of the embodiments of the present invention is not limited thereto.

Referring to FIG. 1 and FIG. 2 , this embodiment provides a single-stream single-stage network target tracking method based on cascaded attention, and the method comprises the following steps:

Step 1: In a single-stream single-stage framework, based on the Transformer network model and the feature enhancement module, a backbone feature extraction and fusion module is constructed. The backbone feature extraction and fusion module, the head corner module, and the fractional head prediction module constitute a single-stream single-stage overall model;

Step 2: Obtain a template image and a search image, wherein the template image includes an initial template containing a number of target objects to be tracked and a number of online templates containing target states;

Step 3: Input the template image and the search image into the single-stream single-stage overall model, and extract the local feature information corresponding to the template image and the search image through the backbone feature extraction and fusion module;

The local feature information is input into the feature enhancement module, and the local semantic information is aggregated using cascaded attention to achieve feature enhancement, thus obtaining the global context information of the template image and the search image.

Perform cross-attention calculation on the global context information of the template image and the search image to achieve communication and obtain the resulting feature map;

In step 3, the template image and the search image are input into the single-stream single-stage overall model, and the method of extracting local feature information corresponding to the template image and the search image through the backbone feature extraction and fusion module specifically includes the following steps:

For each template image, its initial shape is First, we perform word embedding operation on it, and then perform convolution operation to get the result, which has the shape of /> , and then stretch them from two dimensions to one dimension, that is, reshape them into , hereby order/> , where /> Indicates the length of the template token, /> Respectively represent the length, width, and number of channels of the initial template image. /> Respectively represent the length, width, and number of channels of the template image after convolution;

For each search image, its initial shape is The word embedding operation is also performed on it, and then the convolution operation is performed to obtain the result, which has the shape of /> shape, and then stretch them from two dimensions to one dimension, that is, reshape them into/> , hereby order/> , where /> Indicates the length of the search token, /> Respectively represent the length, width, and number of channels of the initial search image. /> Respectively represent the length, width, and number of channels of the search image after convolution, and/> ,/> ,/> .

Please refer to FIG. 3 and FIG. 4. In step 3, the method of aggregating local semantic information by using cascade attention to achieve feature enhancement and obtaining global context information of the template image and the search image specifically includes the following steps:

The total input token includes template tokens and search tokens. Template tokens include an initial template token and several online template tokens. The template tokens and search tokens are concatenated. The concatenation process has the following relationship:

;

in, Represents the concatenation function, which by default performs concatenation according to the channel dimension, with the purpose of concatenating multiple tokens to obtain the final input./> Represents the total input tokens, /> represents the initial template token among the total input tokens, Represents several online template tokens in the total input tokens, and its shape is/> ,/> Represents the search token in the total input tokens, and its shape is /> ;

Convert the total input tokens into a two-dimensional image. The conversion process has the following relationship:

;

in, Indicates the first/> A two-dimensional image, /> Indicates the first/> input tokens, /> Represents a function that transforms a one-dimensional vector into a two-dimensional image; for example, the shape of X is/> , after changing the function, the shape becomes/> , where; /> ;

Then the two-dimensional image is input into the self-attention enhancement function for feature extraction to obtain the enhanced token corresponding to each image. The process of feature extraction using the self-attention enhancement function has the following relationship:

;

in, Represents the Self Attention Enhancement Module, Indicates the first/> Enhancement tokens;

The enhanced tokens of the template image part are connected. The process of enhancing the token connection has the following relationship:

;

in, The template result token obtained by concatenating the tokens representing the template image part, /> Represents a concatenation function, which by default performs concatenation according to the channel dimension, with the goal of concatenating multiple tokens to obtain the final template result token.

In step 3, cross attention calculation is performed on the global context information of the template image and the search image to achieve communication, and the method for obtaining the result feature map specifically includes the following steps:

The query, key, and value are generated using the template result token and the enhanced token corresponding to the search image. The generation process of query, key, and value has the following relationship:

;

;

;

in, Represents the query, key, value about the enhanced token, /> denote the convolution operations on the query, key, and value of the enhanced token, respectively;

Cross-attention calculation is performed on query, key, and value. The cross-attention calculation process has the following relationship:

;

in, represents the search token after cross attention, /> Indicates the dimension of the key, /> Indicates the matrix transposed state, /> represents the softmax function used to calculate the attention weights to transform the original attention scores into probability distributions, ensuring that the weights of all positions are between 0 and 1 and their sum is equal to 1;

The cross attention calculation result is concatenated with the template result token to obtain the total token. The process of concatenating the cross attention calculation result with the template result token has the following relationship:

;

in, represents the total tokens after cross attention;

The total tokens are passed through layer normalization and multi-layer perceptron in turn to obtain the resulting feature map. The process of passing the total tokens through layer normalization and multi-layer perceptron in turn has the following relationship:

;

;

in, Indicates the temporary storage result status, /> Indicates the output of the current calculation part, /> represents a multilayer perceptron, /> Represents the layer normalization function (Layer Norm).

Furthermore, the two-dimensional image is input into the self-attention enhancement function for feature extraction, and the method for obtaining the enhanced token corresponding to each image specifically includes the following steps:

The first A two-dimensional image/> and the /> Attention heads output tokens /> Add as the new /> A two-dimensional image/> , the process expression is as follows:

;

in, Indicates the first/> Attention head, /> Indicates the new /> A two-dimensional image;

Using self-attention method, A two-dimensional image/> As new input to calculate the new / > Attention head, denoted as/> , the process expression is as follows:

;

;

in, represents the self-attention function,/> Respectively represent the convolution operations on the query, key, and value of the i -th enhanced token, /> represents the dimension of the key;

After connecting the outputs of all new attention heads in the form of convolution operations, we apply normal convolution operations to enhance the local information of the features and obtain the enhanced tokens corresponding to each image. The process expression is as follows:

;

in, Represents a normal convolution operation.

In this step, A two-dimensional image/> and the /> Attention heads output tokens /> Add as the new /> A two-dimensional image/> , using the self-attention method to A two-dimensional image/> As new input to calculate the new / > After connecting all the heads in the form of convolution operation, we also apply /> That is, ordinary convolution operations are used to enhance the local information of features. This enables the self-attention mechanism to fully capture local and global relationships and further enhance feature representation.

Step 4: Segment the result feature map into a template image and a search image, and use them as the input for the next stage. Repeat step 3 several times in an iterative manner to obtain the final result feature map.

Step 5: Input the final result feature map into the head corner point module to predict the confidence score of each target position, and determine the position of the tracking target according to the confidence score to achieve target tracking;

The resulting feature map is input into the score head prediction module to predict the confidence score of each target state, and the confidence score of the target state is used to determine whether to use the predicted target state as an online template in the next stage of online tracking.

The result feature map is input into the score head prediction module to predict the confidence score of each target state. The method of determining whether to use the predicted target state as an online template in the next stage of online tracking process according to the confidence score of the target state specifically includes the following steps:

The fraction token that will be learned Generate a query to search for region of interest tokens. The process expression is as follows:

;

in, Expressing support for/> The query convolution operation, /> Represents a query to search for region of interest tokens, /> represents a learned fraction token, where the shape is 1×C;

Then, the important regional features are adaptively extracted from the result feature map, and the key and value are generated using the important regional features. The process expression is as follows:

;

;

;

in, Represents the result feature map, /> 、/> Respectively represent the key and value of important regional features,/> Represents the ROI function used to adaptively extract important regional features,/> Indicates important regional features, /> Represents the key convolution operation on important regional features,/> Represents the value convolution operation on the important regional features;

Then, the attention weight is calculated using the query, key, and value, and the attention weight is passed through the multi-layer perceptron and the Sigmoid activation function in turn to obtain the prediction score. The process expression is as follows:

;

;

in, represents the attention weight of the attention output,/> represents the activation function used to generate a score from 0 to 1, /> represents the confidence score;

When the confidence score is lower than 0.5, the online template will be regarded as negative, indicating that it will not be updated. Otherwise, it will be regarded as positive, indicating that the online template will be updated and the predicted target state will be used as the online template in the next stage of online tracking process.

In this step, the fractional tokens that can be learned is used as a query to participate in the search for ROI tokens, so that the score token can encode the mined target information. Next, the score token focuses on all positions of the initial target token to implicitly compare the mined target with the first target. Finally, the score is generated by the MLP layer and the Sigmoid activation function and the online template is updated according to the score. In this way, we are able to effectively screen and update the online template and improve the accuracy and stability of the tracking system.

Step 6: Repeat steps 2 to 4 based on a large-scale dataset to pre-train the single-stream single-stage overall model to optimize model parameters.

Step 7: Use the trained single-stream single-stage overall model to track the target online in the video sequence.

Referring to FIG. 5 , this embodiment further provides a single-stream single-stage network target tracking system based on cascaded attention, wherein the system applies the single-stream single-stage network target tracking method based on cascaded attention as described above, and the system includes:

Building blocks for:

In the single-stream single-stage framework, based on the Transformer network model and the feature enhancement module, a backbone feature extraction and fusion module is constructed. The backbone feature extraction and fusion module, the head corner module, and the fractional head prediction module constitute the single-stream single-stage overall model.

Learning modules for:

Acquire a template image and a search image, wherein the template image includes an initial template containing a number of desired tracking targets and a number of online templates containing target states;

The template image and the search image are input into the single-stream single-stage overall model, and the local feature information corresponding to the template image and the search image is extracted through the backbone feature extraction and fusion module;

The local feature information is input into the feature enhancement module, and the local semantic information is aggregated using cascaded attention to achieve feature enhancement, and the global context information of the template image and the search image is obtained;

Perform cross-attention calculation on the global context information of the template image and the search image to achieve communication and obtain the resulting feature map;

Extraction module for:

The resulting feature map is divided into a template image and a search image, and used as the input for the next stage. The feature extraction is repeated several times in an iterative manner to obtain the final result feature map.

Compute module for:

The final result feature map is input into the head corner module to predict the confidence score of each target position, and the position of the tracking target is determined according to the confidence score to achieve target tracking;

The resulting feature map is input into the score head prediction module to predict the confidence score of each target state, and the confidence score of the target state is used to determine whether to use the predicted target state as an online template in the next stage of online tracking.

Pre-trained modules for:

Pre-train the single-stream single-stage overall model based on large-scale datasets to optimize model parameters;

Tracking module for:

The trained single-stream single-stage holistic model is used to perform online object tracking in video sequences.

In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims (2)

1.A method for single-stream single-stage network target tracking based on cascade attention, the method comprising the steps of:

Step 1, under a single-stream single-stage framework, constructing and obtaining a trunk feature extraction and fusion module based on a Transformer network model and a feature enhancement module, wherein the trunk feature extraction and fusion module, a head corner module and a fractional head prediction module form a single-stream single-stage integral model;

Step 2, obtaining a template image and searching pictures, wherein the template image comprises an initial template containing a plurality of required tracking targets and a plurality of online templates containing target states;

Step 3, inputting the template image and the search picture into a single-stream single-stage integral model, and extracting local feature information corresponding to the template image and the search picture through a trunk feature extraction and fusion module;

inputting the local feature information into a feature enhancement module, and aggregating the local semantic information by using cascade attention to realize feature enhancement, so as to obtain global context information of a template image and a search picture;

Performing cross attention calculation on global context information of the template image and the search picture to realize communication, and obtaining a result feature map;

Step 4, dividing the result feature map into a template image and a search picture, and repeating the step 3 for a plurality of times in an iterative mode to obtain a final result feature map;

step 5, inputting the final result feature map into a head corner module to predict the confidence score of each target position, and determining the position of the tracking target according to the confidence score so as to realize target tracking;

inputting the result feature map into a score head prediction module to predict the confidence score of each target state, and determining whether to take the predicted target state as an online template in the online tracking process of the next stage according to the confidence score of the target state;

step 6, repeating the steps 2 to 4 by using the large-scale data set as a basis, and pre-training the single-flow single-stage integral model to optimize model parameters;

Step 7, performing target online tracking on the video sequence by using the trained single-stream single-stage integral model;

In the step 3, the template image and the search picture are input into a single-stream single-stage integral model, and the method for extracting the local feature information corresponding to the template image and the search picture through the trunk feature extraction and fusion module specifically comprises the following steps:

for each template image, its initial shape is Firstly, word embedding operation is carried out on the shape, and then convolution operation is carried out to obtain a result, wherein the shape is/>They are then stretched from two dimensions to one dimension, i.e. remoldedLet/>, hereWherein/>Representing the length of the template token,/>Respectively representing the length, width and channel number of the initial template image,/>Respectively representing the length, width and channel number of the template image after convolution;

for each search image, its initial shape is Similarly, word embedding operation is performed on the obtained product, and convolution operation is performed to obtain a result, wherein the shape is/>They are then stretched from two dimensions to one dimension, i.e. remolded/>Let/>, hereWherein/>Representing the length of the search token,/>Respectively representing the length, width and channel number of the initial search image,/>Respectively representing the length, width and channel number of the convolved search template image, and/>，/>，/>；

In the step 3, the method for aggregating the local semantic information by using the cascade attention to realize feature enhancement and obtaining the global context information of the template image and the search picture specifically comprises the following steps:

The total input tokens comprise template tokens and search tokens, the template tokens comprise an initial template token and a plurality of online template tokens, the template tokens and the search tokens are spliced, and the splicing process has the following relational expression:

；

wherein, Representing a stitching function,/>Representing the total input token,/>Representing an initial template token in the total input token,/>Representing several online template tokens in the total input token, whose shape is/>，/>Representing a search token of the total input tokens, the shape of which is/>；

The total input token is converted into a two-dimensional image, and the conversion process has the following relation:

；

wherein, Represents the/>Two-dimensional image,/>Represents the/>Input tokens,/>A function representing converting the one-dimensional vector into a two-dimensional picture;

Inputting the two-dimensional images into the self-attention enhancing function to perform feature extraction, obtaining an enhancing token corresponding to each image, wherein the process of performing feature extraction by using the self-attention enhancing function has the following relation:

；

wherein, Representing a self-attention enhancing function,/>Represents the/>A plurality of enhancement tokens;

Connecting the enhancement tokens related to the template image part, wherein the enhancement token connection process has the following relation:

；

wherein, Template result token obtained by token stitching representing a template image portion,/>Representing a splicing function;

The method for carrying out cross attention calculation on global context information of the template image and the search picture to realize communication, and obtaining the result feature map specifically comprises the following steps:

generating a query, a key and a value by using the template result token and the enhanced token corresponding to the search image, wherein the generation process of the query, the key and the value has the following relational expression

；

；

；

Wherein,Representing queries, keys, values, and/or the like about enhanced tokensRespectively representing convolution operations on queries, keys, values about the enhanced token;

The cross attention calculation is carried out on the query, the key and the value, and the cross attention calculation process has the following relation:

；

wherein, Representing search tokens after cross-attention,/>Representing the dimensions of the key,/>Represents the transposed state of the matrix,Representing a softmax function for calculating the attention weights to convert the original attention score into a probability distribution, ensuring that the weights for all locations are between 0 and 1 and that their sum equals 1;

splicing the cross attention computing result and the template result token to obtain a total token, wherein the process of splicing the cross attention computing result and the template result token has the following relational expression:

；

wherein, Representing the total token after cross-attention;

sequentially passing the total tokens through the layer normalization and the multi-layer perceptron to obtain a result feature map, wherein the process of sequentially passing the total tokens through the layer normalization and the multi-layer perceptron has the following relational expression:

；

；

wherein, Representing the status of temporary storage results,/>Representing the output of the currently calculated part,/>Representing a multi-layer perceptron,/>Representing a layer normalization function;

The method for inputting the two-dimensional images into the self-attention enhancement function to extract the characteristics and obtaining the enhancement tokens corresponding to each image specifically comprises the following steps:

Will be the first Two-dimensional image/>And/>Individual attention header output tokens/>Add as new/>Two-dimensional image/>The process expression is as follows:

；

wherein, Represents the/>Attention header,/>Represents a new/>A two-dimensional image;

Self-attentive mode is adopted to make the first Two-dimensional image/>Calculate a new/>, as a new inputAttention heads, hereinafter referred to as/>The process expression is as follows:

；

；

wherein, Representing a self-attention function,/>Respectively representing convolution operations on queries, keys, values with respect to the ith enhanced token,/>Representing the dimensions of the key;

after the output of all new attention heads is connected in a convolution operation mode, the common convolution operation is applied to strengthen the local information of the features, and an enhanced token corresponding to each image is obtained, wherein the process expression is as follows:

；

wherein, Representing a normal convolution operation;

in the step 5, the result feature map is input into a score head prediction module to predict the confidence score of each target state, and the method for determining whether to take the predicted target state as an online template in the next stage online tracking process according to the confidence score of the target state specifically comprises the following steps:

score tokens to be learned A query participating in searching for a region of interest token is generated, the process expression of which is as follows:

；

wherein, Representation pair/>Is a query convolution operation,/>Representing a query about searching for a region of interest token,/>A score token representing learning, wherein the shape is 1 xc;

And extracting important region features from the result feature map in a self-adaptive manner, and generating keys and values by using the important region features, wherein the process expression is as follows:

；

；

；

wherein, Representing the resulting feature map,/>、/>Respectively representing characteristic keys and values related to important areas,/>Representing an ROI function for adaptively extracting important region features,/>Representing important regional features,/>Key convolution operations representing features of important regions,/>A value convolution operation representing a feature of the region of interest;

And calculating attention weight by using query, key and value, and sequentially passing the attention weight through a multi-layer perceptron and a Sigmoid activation function to obtain a prediction score, wherein the process expression is as follows:

；

；

wherein, Attention weight representing attention output,/>Representing an activation function for generating a 0 to 1 score,/>Representing a confidence score;

When the confidence score is lower than 0.5, the online template is considered negative, which indicates that the online template is not updated, otherwise, the online template is considered positive, which indicates that the online template is updated, and the predicted target state is taken as the online template in the online tracking process of the next stage.

2. A cascade attention-based single-flow single-phase network object tracking system, wherein the system applies the cascade attention-based single-flow single-phase network object tracking method as claimed in claim 1, the system comprising:

A construction module for:

Under a single-flow single-stage framework, constructing and obtaining a trunk feature extraction and fusion module based on a Transformer network model and a feature enhancement module, wherein the trunk feature extraction and fusion module, the head corner module and the fractional head prediction module form a single-flow single-stage integral model;

a learning module for:

acquiring a template image and a search picture, wherein the template image comprises an initial template containing a plurality of required tracking targets and a plurality of online templates containing target states;

inputting the template image and the search picture into a single-stream single-stage integral model, and extracting local feature information corresponding to the template image and the search picture through a trunk feature extraction and fusion module;

inputting the local feature information into a feature enhancement module, and aggregating the local semantic information by using cascade attention to realize feature enhancement, so as to obtain global context information of a template image and a search picture;

Performing cross attention calculation on global context information of the template image and the search picture to realize communication, and obtaining a result feature map;

An extraction module for:

dividing the result feature map into a template image and a search picture, and repeating feature extraction for a plurality of times in an iterative mode to obtain a final result feature map;

A calculation module for:

inputting the final result feature map into a head corner module to predict the confidence score of each target position, and determining the position of a tracking target according to the confidence score so as to realize target tracking;

inputting the result feature map into a score head prediction module to predict the confidence score of each target state, and determining whether to take the predicted target state as an online template in the online tracking process of the next stage according to the confidence score of the target state;

a pre-training module for:

Pre-training the single-flow single-stage integral model by using the large-scale data set as a base to optimize model parameters;

A tracking module for:

and carrying out target online tracking on the video sequence by utilizing the trained single-stream single-stage integral model.