A target recognition and positioning system based on image segmentation and its terminal device

Abstract

The present invention relates to the field of target recognition technology, specifically a target recognition and positioning system based on image segmentation and its terminal device, the system includes an edge processing module, a feature recognition module, a regression analysis module, an attention weighting module, an abnormal analysis module, and a target positioning module. In the present invention, the recognition ability of the target in the image is enhanced by using texture, shape and color analysis, the accuracy and reliability of recognition are improved, the prediction of the target position is refined by probability analysis and regression technology, the positioning error is reduced, the positioning accuracy is improved, the dynamic weight adjustment highlights the important features of the target area, and the focus of target identification is optimized. The abnormal analysis calculates the deviation between the target and the predetermined coordinates, enhances the system's recognition and response capabilities to abnormal situations, improves the adaptability and safety of the system, and jointly improves the overall performance of the target recognition and positioning system and the practicality of application.

Description

Technical Field

[0001] The present invention relates to the technical field of target recognition, and in particular to a target recognition and positioning system based on image segmentation and a terminal device thereof. Background Art

[0002] Object recognition is a computer vision technology that focuses on detecting, identifying, and locating entities from images or videos. It identifies various targets such as people, vehicles, or buildings by analyzing the content of digital images. Object recognition technology is widely used in security monitoring, autonomous driving, industrial automation, and daily applications such as smart photo albums and social media to help improve image search and content screening. The research involves multiple sub-fields such as image processing, machine learning, and pattern recognition, and is an important direction in artificial intelligence research.

[0003] Among them, the target recognition and positioning system based on image segmentation refers to a high-precision target recognition technology that can identify and locate specific targets in complex backgrounds. Advanced image segmentation technology is used to improve the accuracy of recognition by accurately separating the foreground and background in the image. It has a wide range of uses. For example, in the military and security fields, it can be used to identify and track enemy targets. In the medical field, it can help accurately locate diseased tissues. In robotics, it enables robots to more accurately identify and operate objects.

[0004] Although existing target recognition technologies have been widely used in many fields, they face the problems of insufficient recognition accuracy and high misrecognition rate when dealing with target recognition and positioning in complex backgrounds. In particular, it is difficult for existing technologies to accurately identify and locate targets when the environment changes greatly or the target features are not obvious. For example, in security monitoring, complex backgrounds cause important targets to be ignored or mislabeled. When processing dynamic images, existing technologies cannot accurately track moving targets due to the lack of effective dynamic analysis and adaptation mechanisms, resulting in impaired real-time and accuracy. For example, in autonomous driving systems, it leads to misjudgment of the surrounding environment, which in turn affects decision-making and safety. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a target recognition and positioning system based on image segmentation and a terminal device thereof.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a target recognition and positioning system based on image segmentation includes:

[0007] The edge processing module is based on the image data, adjusts the contrast and brightness of the image, sharpens the image edge, extracts the brightness and color histogram of the image, and obtains the processed image data;

[0008] The feature recognition module uses the processed image data to analyze the texture, shape and color information in the image, recognizes local and global features from the multi-layer feature map, and constructs an image feature vector;

[0009] The regression analysis module uses the image feature vector to perform probability analysis on the position of the pixel point, perform target recognition, optimize the error of the coordinate point prediction, and obtain the target coordinate prediction result;

[0010] The attention weighting module implements dynamic weight adjustment according to the target coordinate prediction result, focuses on analyzing the characteristics of the target area, and obtains the optimized coordinate position by adjusting the weight of each feature;

[0011] The abnormality analysis module uses the optimized coordinate position to analyze the deviation between the target and the predetermined normal coordinates, calculates the degree of deviation between the target and the preset area, and marks the abnormal target area to obtain the abnormality detection result;

[0012] The target positioning module identifies the abnormal cause of the target positioning based on the abnormal detection result, verifies the target coordinates by analyzing and comparing the positioning data of the differentiated targets, and generates positioning and recognition results.

[0013] As a further solution of the present invention, the processed image data includes an image that has been subjected to contrast and brightness adjustment and edge sharpening processing, and a brightness and color histogram of the image, the image feature vector includes a vector of a multi-level feature map constructed from texture, shape and color information, the target coordinate prediction result includes a pixel point position and a prediction error optimized by probability analysis and target recognition, the optimized coordinate position includes a coordinate point position adjusted by dynamic weight adjustment and target area feature analysis, the abnormality detection result includes a marked abnormal target area and a degree of deviation between the calculated target and a preset area, and the positioning and recognition result includes output data of abnormal cause analysis through target positioning and differentiated target coordinate verification.

[0014] As a further solution of the present invention, the edge processing module includes:

[0015] The contrast adjustment submodule uses image data, subdivides the image, adjusts color parameters to optimize color levels, adjusts brightness parameters to identify image edges, optimizes image contrast, and obtains a contrast-optimized image;

[0016] The histogram extraction submodule applies histogram equalization processing to adjust the image brightness and color distribution based on the contrast optimized image, extracts the brightness and color histograms of the image, and obtains histogram data;

[0017] The sharpening processing submodule identifies the image edge based on the histogram data, optimizes the clarity of the edge area through local contrast, evaluates the overall visual effect of the image, and obtains processed image data.

[0018] As a further solution of the present invention, the feature recognition module includes:

[0019] The texture analysis submodule uses the processed image data to analyze the texture pattern and repeated structure in the image, records the local texture features of the target in the image, and generates texture feature data;

[0020] The shape recognition submodule uses the texture feature data to recognize and classify various shapes in the image, including circles, rectangles and irregular shapes, records global and local shape information, and obtains shape feature vectors;

[0021] The color analysis submodule analyzes the color distribution and the spatial relationship of colors in the image based on the shape feature vector and the color histogram, identifies key color information from multi-layer feature maps, and constructs an image feature vector.

[0022] As a further solution of the present invention, the regression analysis module includes:

[0023] The probability calculation submodule uses the image feature vector to perform statistical analysis on the position distribution of pixels in the image, calculates the probability that each pixel belongs to the target object, and generates pixel probability analysis results by quantifying the distribution and density of the pixels;

[0024] The feature matching submodule identifies the target in the image based on the pixel probability analysis result, matches the features of the differentiated target shape and color, and identifies the key features to obtain target recognition information;

[0025] The coordinate optimization submodule uses the target recognition information to optimize the prediction error of the coordinate points according to differentiated image conditions and target types to obtain the target coordinate prediction result.

[0026] As a further solution of the present invention, the attention weighting module includes:

[0027] The target feature analysis submodule uses the target coordinate prediction data to extract spatial features and time series data in the target area, analyze key features and screen the affected feature set, analyze and process the features, and generate key morphological features;

[0028] The dynamic weight adjustment submodule uses the key morphological features to initialize the weight allocation according to the spatial and temporal correlation of the features, adjusts the weight ratio in combination with the dynamic changes of the target, optimizes the weight allocation mechanism, verifies the dynamic adaptability of the weight, and obtains the dynamic prediction analysis results of the offset;

[0029] The coordinate calculation submodule uses the offset dynamic prediction analysis results to iteratively calculate the coordinate data, performs coordinate correction according to the prediction error, verifies the consistency of the output results, and obtains the optimized coordinate position.

[0030] As a further solution of the present invention, the abnormality analysis module includes:

[0031] The deviation calculation submodule adopts the optimized coordinate position and the Euclidean distance algorithm to compare the real-time coordinates of the target with the predetermined normal coordinates, calculates the Euclidean distance between the coordinate points, determines the degree of deviation through statistical analysis, and generates deviation degree data;

[0032] The abnormal marking submodule identifies the coordinate points beyond the normal range through the deviation degree data, marks the coordinate points as abnormal, and prioritizes the abnormal areas to obtain abnormal target area data;

[0033] The abnormal association submodule identifies target abnormal association information in graphic and text form based on the abnormal target area data and in combination with abnormal coordinate point information, and checks the consistency of the information to obtain abnormal detection results.

[0034] As a further solution of the present invention, the formula of the Euclidean distance algorithm is as follows:

[0035] ;

[0036] in, is the Euclidean distance between the target real-time coordinates and the predetermined normal coordinates, and is the real-time coordinate point of the target, and is the predetermined normal coordinate point, and is the weight coefficient of the coordinate point x and y, and is the height coordinate point, is the weight coefficient of the height coordinate point, and is the time coordinate point, is the weight coefficient of the time coordinate point.

[0037] As a further solution of the present invention, the target positioning module includes:

[0038] The cause identification submodule uses the abnormal detection results to analyze the causes of abnormal targets, identifies potential influencing factors based on the comparative analysis of abnormal coordinates and normal coordinates, and generates abnormal cause analysis results;

[0039] The data comparison submodule compares the positioning data of the differentiated targets through the abnormal cause analysis results, checks the differences and consistency between the data, and obtains the differentiated data comparison results;

[0040] The coordinate verification submodule uses the differential data comparison result to check and correct the identified target positioning data, verify the consistency and synchronization of the target positioning data with the real-time data, and obtain the positioning and recognition results.

[0041] A target recognition and positioning terminal based on image segmentation comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the target recognition and positioning system based on image segmentation is realized.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are:

[0043] In the present invention, the clarity and detail of the image are improved by adjusting the contrast and brightness and sharpening the image edges, making the feature extraction more accurate and detailed. Texture, shape and color analysis are used to enhance the recognition ability of the target in the image, and the accuracy and reliability of the recognition are improved. Through probability analysis and regression technology, the prediction of the target position is refined, the positioning error is reduced, and the positioning accuracy is improved. Dynamic weight adjustment highlights the important features of the target area, optimizes the focus of target identification, and further enhances the system's recognition and positioning capabilities for the target. Abnormal analysis calculates the deviation between the target and the predetermined coordinates, enhances the system's recognition and response capabilities for abnormal situations, improves the system's adaptability and safety, and jointly improves the overall performance of the target recognition and positioning system and the practicality of the application. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a system flow chart of the present invention;

[0045] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0046] Figure 3 is a flow chart of the edge processing module of the present invention;

[0047] Figure 4 It is a flow chart of the feature recognition module of the present invention;

[0048] Figure 5 It is a flow chart of the regression analysis module of the present invention;

[0049] Figure 6 This is a flow chart of the attention weighting module of the present invention;

[0050] Figure 7 This is a flow chart of the abnormality analysis module of the present invention;

[0051] Figure 8 This is a flow chart of the target positioning module of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention 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 invention and are not intended to limit the present invention.

[0053] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0054] See also Figure 1 to Figure 2 The present invention provides a technical solution, a target recognition and positioning system based on image segmentation, comprising:

[0055] The edge processing module is based on the image data, and sharpens the image edge by adjusting the contrast and brightness of the image, extracts the brightness and color histogram of the image, and adjusts the histogram contrast to optimize the image, thereby obtaining the processed image data;

[0056] The feature recognition module uses the processed image data to analyze the texture, shape and color information in the image through a deep convolutional network, identifies local and global features from multi-layer feature maps, and constructs an image feature vector;

[0057] The regression analysis module uses the image feature vector to refine the features to the pixel-level location information, performs probability analysis on the pixel location information, performs target recognition, optimizes the error of coordinate point prediction, and obtains the target coordinate prediction result;

[0058] The attention weighting module implements dynamic weight adjustment based on the target coordinate prediction results, focusing on analyzing the characteristics of the target area, and obtains the optimized coordinate position by adjusting the weight of each feature;

[0059] The abnormality analysis module uses the optimized coordinate position to analyze the deviation between the target area and the predetermined normal coordinates, calculates the degree of deviation between the target area and the preset area, determines whether the target is abnormal, and marks the abnormal target area to obtain the abnormality detection result;

[0060] The target positioning module identifies the abnormal causes of target positioning based on the abnormal detection results, verifies the target coordinates by analyzing and comparing the positioning data of differentiated targets, and generates positioning and recognition results.

[0061] The processed image data includes images that have been adjusted for contrast and brightness, sharpened for edges, and brightness and color histograms of the images. The image feature vector includes a vector of a multi-level feature map constructed from texture, shape, and color information. The target coordinate prediction results include pixel point positions and prediction errors optimized through probability analysis and target recognition. The optimized coordinate positions include those adjusted through dynamic weight adjustment, target area feature analysis, and coordinate point positions. The anomaly detection results include the marked abnormal target area and the degree of deviation between the calculated target and the preset area. The positioning and identification results include output data of abnormal cause analysis through target positioning and differentiated target coordinate verification.

[0062] See also Figure 2 , 3 , the edge processing module includes:

[0063] The contrast adjustment submodule uses image data, adjusts color parameters to optimize color levels by subdividing the image, adjusts brightness parameters to identify image edges, optimizes the image contrast, and obtains the contrast-optimized image. The execution process is as follows;

[0064] The image data is segmented and the color and brightness parameters are adjusted to optimize the contrast of the image. In the process, the adjustment of color parameters is aimed at enhancing the color layering of the image, making the colors more vivid and distinct. Brightness adjustment is mainly used to improve the recognition ability of image edges and highlight details by enhancing the light and dark contrast of image edges. The adjustment makes the visual effect of the image more vivid and helps subsequent image processing such as feature extraction and target recognition. Contrast optimization increases the interpretability of image information, which is especially important when processing images taken under low light conditions. To obtain a contrast-optimized image, the formula used is:

[0065] ;

[0066] in, represents the adjusted contrast image, is the brightness level of the original image, is the brightness adjustment factor, are color adjustment parameters.

[0067] The histogram extraction submodule optimizes the image based on contrast, applies histogram equalization to adjust the image brightness and color distribution, extracts the image brightness and color histograms, and obtains the execution flow of the histogram data as follows;

[0068] Based on the contrast-optimized image, the histogram equalization technique is used to adjust the brightness and color distribution of the image. Histogram equalization is an image processing technique used to improve the global contrast of an image, especially when the available data of the image is limited due to backlight or insufficient illumination. By effectively stretching the brightness distribution of the image, the entire image can obtain uniform brightness. It not only enhances the visual effect of the image, but also improves the accuracy of image analysis, such as in pattern recognition and computer vision tasks. By extracting the brightness and color histograms of the image, detailed data about the color distribution of the image can be obtained to obtain histogram data.

[0069] The sharpening processing submodule identifies the image edge based on the histogram data, optimizes the clarity of the edge area through local contrast, evaluates the overall visual effect of the image, and obtains the processed image data. The execution process is as follows;

[0070] Based on the information obtained from the histogram data, the image edges are identified and enhanced, the local contrast of the image is improved, and the clarity of the edge area is optimized. Through sharpening processing, the details of the image can be made clearer, especially at the edge of the image, which is critical to improving the visual quality of the image and subsequent image analysis. Sharpening is achieved by increasing the brightness difference of the edge pixels of the image, strengthening the high-frequency information in the image, making the image look clearer and more detailed, and obtaining the processed image data.

[0071] See also Figure 2 , 4 , the feature recognition module includes:

[0072] The texture analysis submodule uses the processed image data to analyze the texture patterns and repeated structures in the image, records the local texture features of the target in the image, and generates the texture feature data. The execution process is as follows;

[0073] Analyze the texture patterns and repetitive structures in the processed image data, and extract local texture features in the image, including the roughness, directionality, contrast, and uniformity of the texture, through advanced image processing techniques such as gray-level co-occurrence matrix (GLCM) or Gabor filter. Through features, the visual and structural properties of the surface of objects in the image can be described to help identify and classify different objects and scenes. Texture analysis plays an important role in enhancing the content understanding of the image and supporting subsequent image analysis tasks such as object recognition and scene understanding. Through accurate texture analysis, texture feature data is generated, using the formula:

[0074] ;

[0075] in, Represents texture feature data, is the parameter of the dth texture feature, is the corresponding texture response value, is the number of texture features.

[0076] The shape recognition submodule uses texture feature data to identify and classify various shapes in the image, including circles, rectangles, and irregular shapes, and records global and local shape information to obtain the shape feature vector. The execution process is as follows;

[0077] Use texture feature data to identify and classify various shapes in images, including circles, rectangles, and irregular shapes. By calculating the geometric parameters of the shape, such as boundaries, areas, perimeters, etc., and using shape matching algorithms such as Hough transform or contour analysis, various shapes in the image can be identified. Not only global shape information is recorded, but also local shape variations are analyzed in detail. Shape information is extremely critical for understanding image content and structural levels, and also provides necessary data support for subsequent image classification and recognition. The shape feature vector is generated using the formula:

[0078] ;

[0079] in, represents the shape feature vector, are the length and geometric parameters of the f-th shape, is the shape complexity factor, is the number of shapes recognized.

[0080] The color analysis submodule analyzes the color distribution and spatial relationship of the image based on the shape feature vector and the color histogram, identifies key color information from the multi-layer feature map, and constructs the image feature vector. The execution process is as follows;

[0081] Based on the shape feature vector, combined with the color histogram to analyze the color distribution and spatial relationship of the color in the image, using multi-layer feature mapping technology to identify and classify the key color information in the image, we can analyze in detail how the color is distributed in different objects and scenes, as well as the interaction between colors. Color data is very important for understanding image content, enhancing visual effects, and supporting the decision-making of machine vision systems. To construct the image feature vector, the formula used is:

[0082] ;

[0083] in, represents the image feature vector, is the number of pixels in the g-th color category, is the statistical weight of the color, is the number of color categories.

[0084] See also Figure 2 , 5, the regression analysis module includes:

[0085] The probability calculation submodule uses the image feature vector to perform statistical analysis on the position distribution of pixels in the image, calculates the probability that each pixel belongs to the target object, and generates the pixel probability analysis result by quantifying the distribution and density of the pixels. The execution process is as follows;

[0086] Using the image feature vector, we conduct in-depth statistical analysis on the position distribution of each pixel in the image, calculate the probability that each pixel belongs to a specific target object, and perform quantitative processing based on the distribution and density of the pixels. This involves comparing the color, shape, and texture features of each pixel with the known target features, and calculating the probability of belonging to a category based on the degree of feature matching. This not only helps to more accurately identify and locate the target in the subsequent image processing steps, but also provides data support for further analysis of the image, generating pixel probability analysis results, using the formula:

[0087] ;

[0088] in, represents the probability that the pixel belongs to the target, is the feature vector of the pixel, is the mean vector of target features, is the standard deviation of the feature, is the normalization factor and e is the base of the natural logarithm function.

[0089] The feature matching submodule identifies the target in the image based on the pixel probability analysis results. By matching the features of differentiated target shapes and colors and identifying key features, the execution process of obtaining target recognition information is as follows;

[0090] Based on the pixel probability analysis results, the target in the image is identified. By comparing the shape and color features of the differentiated target and matching them with the corresponding features in the image, the key feature points are identified, and the target in the image can be accurately located and identified. In the feature matching process, the geometric properties of the shape and the distribution characteristics of the color are taken into account to ensure the accuracy and efficiency of the matching process, and the target identification information is obtained. The formula used is:

[0091] ;

[0092] in, represents the target matching score, is the matching score of the z-th target shape, is the color matching score, is the distance between the target feature and the image feature, is the number of targets.

[0093] The coordinate optimization submodule uses the target recognition information to optimize the prediction error of the coordinate points according to the differentiated image conditions and target types, and the execution process of obtaining the target coordinate prediction result is as follows;

[0094] Using target recognition information, we optimize the prediction error of coordinate points for different image conditions and target types. By analyzing the position, size, and changes in the surrounding environment of the target, we can accurately adjust the prediction model, reduce the prediction error, and improve the accuracy of positioning. Coordinate optimization is performed through complex algorithms to ensure that the best prediction results can be obtained under different conditions. It is not only highly accurate, but also has good robustness, adapting to the needs of various complex environments, and generating target coordinate prediction results. The formula used is:

[0095] ;

[0096] in, Represents the coordinate point prediction result, is the importance weight of the zth target, is the adjustment parameter, is the distance between the original prediction and the actual position, is the number of targets, and e is the base of the natural logarithm function.

[0097] See also Figure 2 , 6 , the attention weighted module includes:

[0098] The target feature analysis submodule uses the target coordinate prediction data to extract the spatial features and time series data in the target area, analyzes the key features and screens the affected feature set, analyzes and processes the features, and generates the key morphological features. The execution process is as follows;

[0099] Using the target coordinate prediction data, we deeply extract the spatial features and time series data in the target area, and through careful analysis of the data, we screen out the key feature set that has a significant impact on target identification and tracking. The process involves advanced data processing techniques, such as feature extraction, dimensionality reduction, and feature selection, to ensure that the most valuable information is extracted from complex data. The screened features are further processed, including normalization and standardization, to improve the efficiency and accuracy of subsequent processing steps, and generate key morphological features. The formula used is:

[0100] ;

[0101] in, Represents key morphological features, is the number of features, is the importance weight of feature o, is the numerical value of each feature.

[0102] The dynamic weight adjustment submodule uses key morphological features to initialize and allocate weights according to the spatial and temporal correlation of the features, adjusts the weight ratio in combination with the dynamic changes of the target, optimizes the weight allocation mechanism, verifies the dynamic adaptability of the weight, and obtains the execution process of the offset dynamic prediction analysis results as follows;

[0103] Through key morphological features, weights are initialized and allocated according to the spatial and temporal correlation of the features. Combined with the dynamic changes of the target, the weight ratio is adjusted in real time to optimize the weight allocation mechanism. In the process, the dynamic adaptability of the weights will be verified to ensure that the weight allocation can effectively adapt to the changes in the target state. The optimization of the mechanism helps to improve the response speed and prediction accuracy of the model, especially when the target behavior changes rapidly, and the offset dynamic prediction analysis results are obtained. The formula used is:

[0104] ;

[0105] in, represents the adjusted weight, is the initial weight setting, exp is the natural exponential function, is the weight decay coefficient, is the time variation.

[0106] The coordinate calculation submodule uses the offset dynamic prediction analysis results to iteratively calculate the coordinate data, performs coordinate correction based on the prediction error, verifies the consistency of the output results, and obtains the optimized coordinate position. The execution process is as follows;

[0107] Using the results of the dynamic prediction analysis of the offset, the coordinate data is iteratively calculated to optimize the accuracy of the coordinates. By analyzing the prediction error, necessary coordinate corrections are made to ensure the consistency and accuracy of the output results. Gradually approaching the true position of the target through an iterative algorithm is crucial to ensuring the overall positioning accuracy of the system. The optimization of coordinate calculation, especially when the environment or target conditions change, can effectively improve the stability and reliability of the system, and obtain the optimized coordinate position using the formula:

[0108] ;

[0109] in, represents the optimized coordinate position, is the coordinate position calculated last time, is the target coordinate position, is the learning rate.

[0110] See also Figure 2 , 7 , the exception analysis module includes:

[0111] The deviation calculation submodule adopts the optimized coordinate position, uses the Euclidean distance algorithm, compares the real-time coordinates of the target with the predetermined normal coordinates, calculates the Euclidean distance between the coordinate points, determines the degree of deviation through statistical analysis, and generates the execution process of the degree of deviation data as follows;

[0112] Using the optimized coordinate position, the distance between the target's real-time coordinates and the predetermined normal coordinates is measured using the Euclidean distance algorithm. By calculating the Euclidean distance between coordinate points, the degree of coordinate deviation can be accurately quantified. The process involves a detailed inspection of each coordinate point. Determining the magnitude of coordinate deviation through statistical analysis is essential for evaluating the overall accuracy of the system and the reliability of target positioning. It also provides a basis for subsequent anomaly detection and correction, and generates deviation degree data.

[0113] The formula for the Euclidean distance algorithm is as follows:

[0114] ;

[0115] in, is the Euclidean distance between the target real-time coordinates and the predetermined normal coordinates, and is the real-time coordinate point of the target, and is the predetermined normal coordinate point, and is the weight coefficient of the coordinate point x and y, and is the height coordinate point, is the weight coefficient of the height coordinate point, and is the time coordinate point, is the weight coefficient of the time coordinate point.

[0116] The execution flow is as follows:

[0117] Calculate coordinate differences and , multiplied by the corresponding weight coefficient and , add the height coordinate point and time coordinate points The difference is multiplied by the corresponding weight coefficient and , calculate the sum of squares of weighted difference values, take square root to get deviation degree data, weight coefficient , , and The method for determining is to optimize by minimizing the mean square error between the actual observed deviation and the predetermined deviation.

[0118] The abnormal marking submodule uses the deviation degree data to identify the coordinate points that are beyond the normal range, marks the coordinate points as abnormal, and prioritizes the abnormal areas to obtain the abnormal target area data. The execution process is as follows;

[0119] Through the deviation degree data, coordinate points that are beyond the normal range are identified and abnormally marked. According to the severity of the deviation of the coordinate points, the abnormal areas are prioritized so that the most urgent situations can be handled more effectively. Not only does it enhance the ability to respond to abnormal situations, but also through priority sorting, it ensures that resources can be concentrated on the areas that need the most attention and generates abnormal target area data.

[0120] The abnormal association submodule identifies the target abnormal association information in the form of graphics and text based on the abnormal target area data and the abnormal coordinate point information, and checks the consistency of the information. The execution process of obtaining the abnormal detection result is as follows;

[0121] According to the information of the abnormal target area, combined with the specific data of the abnormal coordinate point, the information related to the target abnormality is displayed in the form of graphics and text. In-depth analysis of the abnormal data, checking the consistency and relevance of the information, ensuring the accuracy and reliability of the abnormal detection results, can clearly identify and understand the background and cause of the abnormality, provide a scientific basis for taking corresponding measures, and obtain abnormal detection results.

[0122] See also Figure 2 , 8 , the target positioning module includes:

[0123] The cause identification submodule uses the anomaly detection results to analyze the causes of the abnormal targets, identifies potential influencing factors based on the comparative analysis of the abnormal coordinates and the normal coordinates, and generates the execution flow of the abnormal cause analysis results as follows;

[0124] Using the anomaly detection results, we conduct an in-depth analysis of the causes of abnormal targets. The analysis process includes comparing the abnormal coordinates with the normal coordinates to identify potential factors that cause deviations, such as environmental changes, equipment failures, or operational errors. The detailed analysis of the factors is of great significance for understanding the root causes of the anomalies, taking targeted measures, and preventing similar problems from happening in the future, and generating abnormal cause analysis results.

[0125] The data comparison submodule compares the positioning data of the differentiated targets through the abnormal cause analysis results, checks the differences and consistency between the data, and obtains the execution process of the differentiated data comparison results as follows;

[0126] Using the results of abnormal cause analysis, the positioning data of differentiated targets are carefully compared, which not only checks the differences between the data, but also verifies the consistency of the data. This is critical for verifying the accuracy of the data processing process, identifying the source of errors in data processing, and optimizing data processing strategies. It helps to improve the system's data processing capabilities and accuracy and obtain differentiated data comparison results.

[0127] The coordinate verification submodule uses the difference data comparison results to check and correct the identified target positioning data, verify the consistency and synchronization of the target positioning data with the real-time data, and the execution process of obtaining the positioning and recognition results is as follows;

[0128] Using the difference data comparison results, the identified target positioning data is verified and corrected. The process includes verifying the consistency and synchronization of the target positioning data with the real-time data to ensure that all data are accurate. In this way, it is crucial to ensure the accuracy and reliability of the system output and obtain positioning and identification results.

[0129] A target recognition and positioning terminal based on image segmentation comprises a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the target recognition and positioning system based on image segmentation is realized.

[0130] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A target recognition and positioning system based on image segmentation, characterized in that: The system comprises: The edge processing module is based on the image data, adjusts the contrast and brightness of the image, sharpens the image edge, extracts the brightness and color histogram of the image, and obtains the processed image data; The feature recognition module uses the processed image data to analyze the texture, shape and color information in the image, recognizes local and global features from the multi-layer feature map, and constructs an image feature vector; The regression analysis module uses the image feature vector to perform distribution probability analysis on the position of the pixel point, performs target recognition, and uses the target recognition result to optimize the error of the coordinate point prediction to obtain the target coordinate prediction result; The attention weighting module implements dynamic weight adjustment according to the target coordinate prediction result, focuses on analyzing the characteristics of the target area, and obtains the optimized coordinate position by adjusting the weight of each feature; The abnormality analysis module uses the optimized coordinate position to analyze the deviation between the target and the predetermined normal coordinates, calculates the degree of deviation between the target and the preset area, and marks the abnormal target area to obtain the abnormality detection result; The target positioning module identifies the abnormal cause of the target positioning based on the abnormal detection result, verifies the target coordinates by analyzing and comparing the positioning data of the differentiated targets, and generates positioning and identification results; The feature recognition module comprises: The texture analysis submodule uses the processed image data to analyze the texture pattern and repeated structure in the image, records the local texture features of the target in the image, and generates texture feature data; The shape recognition submodule uses the texture feature data to recognize and classify various shapes in the image, including circles, rectangles and irregular shapes, records global and local shape information, and obtains shape feature vectors; The color analysis submodule analyzes the color distribution and spatial relationship of the color in the image based on the shape feature vector and the color histogram, identifies key color information from the multi-layer feature map, and constructs an image feature vector; The attention weighting module comprises: The target feature analysis submodule uses the target coordinate prediction data to extract spatial features and time series data in the target area, analyze key features and screen the affected feature set, analyze and process the features, and generate key morphological features; The dynamic weight adjustment submodule uses the key morphological features to initialize the weight allocation according to the spatial and temporal correlation of the features, adjusts the weight ratio in combination with the dynamic changes of the target, optimizes the weight allocation mechanism, verifies the dynamic adaptability of the weight, and obtains the dynamic prediction analysis results of the offset; The coordinate calculation submodule uses the offset dynamic prediction analysis results to iteratively calculate the coordinate data, performs coordinate correction according to the prediction error, verifies the consistency of the output results, and obtains the optimized coordinate position.

2. The target recognition and positioning system based on image segmentation according to claim 1, characterized in that: The processed image data includes an image that has been subjected to contrast and brightness adjustment and edge sharpening processing, and a brightness and color histogram of the image; the image feature vector includes a vector of a multi-level feature map constructed from texture, shape and color information; the target coordinate prediction result includes a pixel point position and a prediction error optimized by probability analysis and target recognition; the optimized coordinate position includes a coordinate point position adjusted by dynamic weight adjustment and target area feature analysis; the anomaly detection result includes a marked abnormal target area and a calculated degree of deviation between the target and the preset area; the positioning and recognition result includes an output data of anomaly cause analysis through target positioning and differentiated target coordinate verification.

3. The target recognition and positioning system based on image segmentation according to claim 1, characterized in that: The edge processing module comprises: The contrast adjustment submodule uses image data, subdivides the image, adjusts color parameters to optimize color levels, adjusts brightness parameters to identify image edges, optimizes image contrast, and obtains a contrast-optimized image; The histogram extraction submodule applies histogram equalization processing to adjust the image brightness and color distribution based on the contrast optimized image, extracts the brightness and color histograms of the image, and obtains histogram data; The sharpening processing submodule identifies the image edge based on the histogram data, optimizes the clarity of the edge area through local contrast, evaluates the overall visual effect of the image, and obtains processed image data.

4. The target recognition and positioning system based on image segmentation according to claim 1, characterized in that: The regression analysis module comprises: The probability calculation submodule uses the image feature vector to perform statistical analysis on the position distribution of pixels in the image, calculates the probability that each pixel belongs to the target object, and generates pixel probability analysis results by quantifying the distribution and density of the pixels; The feature matching submodule identifies the target in the image based on the pixel probability analysis result, matches the features of the differentiated target shape and color, and identifies the key features to obtain target recognition information; The coordinate optimization submodule uses the target recognition information to optimize the prediction error of the coordinate points according to differentiated image conditions and target types to obtain the target coordinate prediction result.

5. The target recognition and positioning system based on image segmentation according to claim 1, characterized in that: The abnormality analysis module comprises: The deviation calculation submodule adopts the optimized coordinate position and the Euclidean distance algorithm to compare the real-time coordinates of the target with the predetermined normal coordinates, calculates the Euclidean distance between the coordinate points, determines the degree of deviation through statistical analysis, and generates deviation degree data; The abnormal marking submodule identifies coordinate points beyond the normal range through the deviation degree data, marks the coordinate points as abnormal, and prioritizes the abnormal areas to obtain abnormal target area data; The abnormal association submodule identifies target abnormal association information in graphic and text form based on the abnormal target area data and in combination with abnormal coordinate point information, and checks the consistency of the information to obtain abnormal detection results.

6. The target recognition and positioning system based on image segmentation according to claim 5, characterized in that: The formula of the Euclidean distance algorithm is as follows: ； in, is the Euclidean distance between the target real-time coordinates and the predetermined normal coordinates, and is the real-time coordinate point of the target, and is the predetermined normal coordinate point, and is the weight coefficient of the coordinate point x and y, and is the height coordinate point, is the weight coefficient of the height coordinate point, and is the time coordinate point, is the weight coefficient of the time coordinate point.

7. The target recognition and positioning system based on image segmentation according to claim 1, characterized in that: The target positioning module comprises: The cause identification submodule uses the abnormal detection results to analyze the causes of abnormal targets, identifies potential influencing factors based on the comparative analysis of abnormal coordinates and normal coordinates, and generates abnormal cause analysis results; The data comparison submodule compares the positioning data of the differentiated targets through the abnormal cause analysis results, checks the differences and consistency between the data, and obtains the differentiated data comparison results; The coordinate verification submodule uses the differential data comparison result to check and correct the identified target positioning data, verify the consistency and synchronization of the target positioning data with the real-time data, and obtain the positioning and recognition results.

8. A target recognition and positioning terminal device based on image segmentation, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the target recognition and positioning system based on image segmentation as described in any one of claims 1 to 7 when executing the computer program.

Images