Intelligent image recognition and analysis system

Abstract

The invention provides an intelligent image recognition and analysis system. The system comprises an image acquisition module, a feature extraction module, a feature database module, a comparison module and an output module, wherein data interaction processing is carried out among the modules; the image acquisition module is used for performing partition acquisition on a to-be-acquired image, numbering each region and then sequentially sending the regions into the feature extraction module according to a numbering sequence; the feature extraction module performs corresponding extraction on high-dimensional features in each numbered image area through wavelet transform and key feature marking modes, and sends the extracted features to the feature database module; the feature database module is used for establishing a feature database and performing fuzzy matching on the extracted high-dimensional features of the target image and the features in the feature database; a feature point control group is established through fuzzy matching, and data of an approximate feature point control group is selected and sent to a comparison module.

Description

Technical Field

[0001] The present invention belongs to the field of recognition and analysis, and in particular relates to an intelligent image recognition and analysis system. Background Art

[0002] At present, the core of the intelligent image recognition and analysis system, as a tool that integrates advanced artificial intelligence technology, is to use deep learning algorithms, especially convolutional neural networks (CNN), to efficiently and accurately recognize and analyze images. The system can simulate the visual processing process of the human brain, extract key features in the image, such as edges, textures, shapes and colors, and then compare them with the feature patterns in the database to achieve a deep understanding and intelligent interpretation of the image content. The system has shown a wide range of application prospects in many fields such as photo recognition, scene analysis, content recommendation, intelligent album management, medical imaging diagnosis, autonomous driving, security monitoring, etc., greatly improving work efficiency and user experience.

[0003] Although intelligent image recognition and analysis systems have made significant progress, existing technologies still have some shortcomings that cannot be ignored. First, data dependency is a key issue. The performance of image recognition algorithms depends largely on the quality and quantity of training data. If the training data is insufficient or biased, the recognition effect of the algorithm may be greatly reduced, resulting in an increase in misjudgments or missed judgments.

[0004] High consumption of computing resources is also a major challenge faced by the system. Image recognition algorithms usually require high-performance computing devices, such as GPUs and large memories, to support large-scale image processing and feature extraction. In some resource-constrained scenarios, the application of this system may face great difficulties, or even fail to achieve real-time processing.

[0005] Limited generalization ability is also a significant disadvantage of existing technologies. Although image recognition algorithms perform well in specific fields and scenarios, their recognition ability may drop significantly when faced with some new and unseen images. This limits the application of the system in certain scenarios that require high flexibility and adaptability.

[0006] Security issues should not be ignored either. Image recognition algorithms may have certain security risks. For example, adversarial samples generated by technologies such as generative adversarial networks (GANs) may cause the algorithm to produce incorrect recognition results. This may not only affect the accuracy of the system, but also bring potential risks to areas that rely on the system to make decisions, such as autonomous driving and medical diagnosis.

[0007] Poor interpretability is also a shortcoming of existing image recognition algorithms. The decision-making process of the algorithm is often a black box, and it is difficult to explain the logic and basis behind it. This may bring certain problems and challenges in some fields that require a high degree of transparency and interpretability, such as medical diagnosis and legal judgment. Summary of the invention

[0008] The present invention proposes an intelligent image recognition and analysis system, which solves the technical problem of efficiently and accurately identifying and analyzing the features of each region in a complex image. Through modular design, the whole process from image acquisition to feature extraction, database matching, comparative analysis and result output is automated, which improves the accuracy and efficiency of image recognition.

[0009] The technical solution of the present invention is implemented as follows: the intelligent image recognition and analysis system includes an image acquisition module, a feature extraction module, a feature database module, a comparison module and an output module for data interactive processing;

[0010] The image acquisition module collects the image to be collected by partitioning, numbers each area and sends them to the feature extraction module in sequence according to the numbering sequence;

[0011] The feature extraction module extracts high-dimensional features in each numbered image region by wavelet transform and key feature marking, and sends the extracted features to the feature database module;

[0012] The feature database module is used to establish a feature database, perform fuzzy matching on the extracted high-dimensional features of the target image and the features in the feature database; establish a feature point control group through fuzzy matching, select approximate feature point control group data and send it to the comparison module;

[0013] The comparison module performs corresponding analysis and comparison on the feature point control group matched by the feature extraction module and the feature database module, and after exporting the classification result of the target image, feeds it back to the output module;

[0014] The output module receives the classification result of the target image, reorganizes and splices the image areas according to the numbers, and then displays them on the display screen.

[0015] Existing technologies often adopt the overall acquisition method, lacking the detailed division and processing of image regions. However, this system uses the image acquisition module to partition the image and number each region, providing a more accurate data basis for subsequent feature extraction and analysis. This partition acquisition method enables the system to capture the detailed features in the image more carefully and improves the accuracy of recognition.

[0016] In terms of feature extraction, the existing technology usually adopts traditional feature extraction methods, such as edge detection, texture analysis, etc. These methods often have poor effects when processing complex images. The system adopts wavelet transform and key feature marking methods to more effectively extract high-dimensional features in images. As a multi-scale analysis method, wavelet transform can capture detailed information of images at different scales, while key feature marking can highlight key information in images, further improving the accuracy and efficiency of feature extraction. In terms of feature database, the existing technology usually uses fixed feature templates for matching, which lacks flexibility and adaptability. The system establishes a feature database through a feature database module and adopts a fuzzy matching method, which can more flexibly handle feature differences between different images. Fuzzy matching improves the accuracy and robustness of matching by establishing a feature point control group and selecting approximate feature point data for comparative analysis. In terms of comparative analysis and result output, the existing technology often lacks a systematic analysis and processing mechanism, resulting in inaccurate or unstable recognition results. The system uses a comparison module to perform corresponding analysis and comparison on the feature point control group, which can more accurately determine the classification results of the image. At the same time, the output module reorganizes and splices the image areas according to the numbers and displays them, making the recognition results more intuitive and easy to understand.

[0017] As a preferred implementation, the feature extraction module includes an image magnification unit, a key area selection unit, a wavelet transform unit and a key feature extraction module. When the image acquisition module extracts the entire image, the wavelet transform unit separates the image into low-frequency and high-frequency parts through convolution; the key area selection unit acquires low-frequency signals, extracts feature vectors in each low-frequency area, and processes the feature vectors accordingly through the key feature extraction module; the image magnification unit selects a magnification factor according to the image resolution, and after determining the magnification factor, amplifies the image; a number is set for each area of ​​the image, and samples are taken in sequence according to the number and sent to the wavelet transform unit.

[0018] As a preferred implementation, when the image magnification unit selects a wavelet transform unit to sample each image magnification area after the area division, the sampling is performed by image stretching before and during the magnification sampling. The sampling steps are: firstly, edge analysis is performed on the image, and an edge extraction algorithm based on image grayscale difference is used to identify the edge of the image; the grayscale difference between the object in the image and the background feature image is set, and the grayscale difference is gradually increased from 0 to extract the target feature point. After the target feature point is successfully extracted, the target feature point is marked and the corresponding grayscale difference is recorded.

[0019] As a preferred implementation, the feature extraction module marks the target feature points. If the grayscale difference is greater than 0 and less than the set a, a target feature point is marked every two points on the edge line, and a marked point is recorded between each two adjacent target feature points; if the grayscale difference is greater than the set a and less than the set b, each interval point on the edge line is marked as a target feature point, and a marked point is recorded between two adjacent target feature points; if the grayscale difference is greater than the set b and less than the set c, a target feature point is marked every 0.5 points on the edge line, and a marked point is recorded between two adjacent target feature points.

[0020] As a preferred embodiment, the marking points are marked by setting colors so that pixels of corresponding colors are marked on the image: when the grayscale difference is greater than 0 and less than the set a, the marking points are displayed in red and the target feature points are displayed in blue; when the grayscale difference is greater than the set a and less than the set b, the marking points are displayed in white and the target feature points are displayed in pink; when the grayscale difference is greater than the set b and less than the set c, the marking points are displayed in black and the target feature points are displayed in light yellow.

[0021] As a preferred implementation, the feature extraction module performs wavelet multi-scale decomposition on the image features: decomposes the image into a low-frequency coefficient and several high-frequency coefficients of different resolutions by wavelet multi-scale, and then extracts the eigenvalues; performs feature conversion on the extracted eigenvalues ​​and the feature points corresponding to the extracted eigenvalues ​​through a feature conversion formula, and combines the obtained feature point data to obtain high-dimensional features.

[0022] After adopting the above technical solution, the beneficial effects of the present invention are: through partition acquisition and numbering processing, the system can capture the detailed features in the image more meticulously, and improve the accuracy and stability of recognition. This processing method is particularly suitable for complex images or images containing multiple targets, and can achieve accurate recognition and analysis of each target. The use of wavelet transform and key feature marking for feature extraction can more effectively extract high-dimensional features in the image, and improve the accuracy and efficiency of feature extraction. This feature extraction method is not only applicable to different types of images, but also can adapt to images of different scales and resolutions, and has strong versatility and adaptability.

[0023] By establishing a feature database and adopting fuzzy matching, the system can more flexibly handle feature differences between different images, improving the accuracy and robustness of matching. This processing method can cope with complex and changeable image recognition scenarios, improving the practicality and reliability of the system. Through the processing of comparative analysis and result output modules, the system can more accurately judge the classification results of the image and display the results in an intuitive way. This processing method not only improves the accuracy of recognition, but also provides users with a better user experience. At the same time, the modular design of the system allows each module to run and upgrade independently, which facilitates the maintenance and expansion of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 It is a flow chart of the system of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Example:

[0028] like Figure 1 As shown in the figure, the intelligent image recognition and analysis system, as an important part of modern artificial intelligence technology, is designed to achieve efficient and accurate recognition and analysis of image content. The system achieves a complete conversion from the original image to the final recognition result through a series of modular data processing processes, including image acquisition, feature extraction, feature database matching, comparative analysis and result output. The following will explain the working principle of the system in detail and demonstrate its workflow through a specific work scenario.

[0029] The image acquisition module is the entrance of the system and is responsible for capturing the image to be processed. In order to improve the accuracy and efficiency of recognition, the module adopts the partition acquisition method to divide the entire image into several small areas and number each area. In this way, each area can be used as an independent processing unit to facilitate subsequent feature extraction and comparative analysis. The collected image data will be sent to the feature extraction module in the order of the number.

[0030] The feature extraction module is one of the cores of the system, responsible for extracting feature information from the image that has a key impact on the recognition result. This module uses wavelet transform and key feature marking to separate the high-frequency and low-frequency information of the wavelet transformed image and extract the texture, edge and other detail features in the image; at the same time, through key feature marking, the significant feature points in the image are marked and extracted. These feature information will be sent to the feature database module for matching.

[0031] The feature database module is responsible for establishing and maintaining a database containing a large amount of known feature information. When receiving feature information sent by the feature extraction module, the module will fuzzy match it with the features in the database. Fuzzy matching is a flexible matching method that allows certain differences between features so that closer feature points can be matched. Through fuzzy matching, the system can establish a feature point control group and select similar feature point control group data to send to the comparison module.

[0032] The comparison module is another core part of the system, responsible for in-depth analysis and comparison of the feature point control group. This module uses machine learning algorithms or deep learning models to classify and identify the feature information in the feature point control group. Through comparison and analysis, the system can obtain the classification results of the target image, such as object category, scene type, etc. These classification results will be fed back to the output module.

[0033] The output module is the output end of the system, responsible for displaying the classification results to the user in an intuitive way. This module receives the classification results sent by the comparison module and reassembles and splices the image areas according to the numbers to restore the complete image. Then, the recognition results and the original image are displayed to the user through output devices such as display screens. In this way, the user can intuitively see the system's recognition results for the image.

[0034] The following will take a specific work scenario as an example to demonstrate the workflow of the intelligent image recognition and analysis system.

[0035] Suppose that on an automated production line, the products on the assembly line need to be classified and identified. These products may include different types and models of mobile phones, tablets and other electronic devices. In order to achieve this goal, an intelligent image recognition and analysis system can be applied to the production line.

[0036] First, the image acquisition module captures images of products on the assembly line. Since products may be in different positions and orientations, images need to be acquired in different areas. For example, the image can be divided into several fixed-size areas and each area is numbered. In this way, each area can be treated as an independent processing unit for subsequent processing.

[0037] Next, the feature extraction module will extract features from each region. This module will use wavelet transform and key feature labeling to extract detailed features such as texture and edges in the image as well as significant feature points. These feature information will be sent to the feature database module for matching.

[0038] In the feature database module, the system will perform fuzzy matching between the extracted feature information and the features in the database. Since the database contains a large amount of known feature information, the feature point control group that is closest to the extracted feature can be found through fuzzy matching. These feature point control groups will be sent to the comparison module for further analysis.

[0039] In the comparison module, the system will conduct in-depth analysis and comparison of the feature point control group. Using machine learning algorithms or deep learning models, the system can classify and identify the feature point control group. For example, it can determine whether the feature point control group belongs to a specific product category or model. Through comparative analysis, the system can obtain the classification result of the target image.

[0040] Finally, the output module will present the classification results to the user in an intuitive way. This module will receive the classification results sent by the comparison module and reassemble and splice the image areas according to the numbers to restore the complete image. Then, the recognition results and the original image will be presented to the user through output devices such as display screens. In this way, the user can intuitively see the system's recognition results of the product.

[0041] In practical applications, intelligent image recognition and analysis systems can also be combined with automated control systems to achieve automatic sorting and classification of products on the production line. For example, when the system recognizes a certain type of product, it can automatically sort it to the corresponding production line or storage area. This can not only improve production efficiency, but also reduce manual intervention and error rates.

[0042] The feature extraction module includes an image magnification unit, a key area selection unit, a wavelet transformation unit and a key feature extraction module. When the image acquisition module extracts the entire image, the wavelet transformation unit separates the image from low frequency to high frequency through convolution; the key area selection unit acquires low-frequency signals, extracts feature vectors in each low-frequency area, and processes the feature vectors accordingly through the key feature extraction module; the image magnification unit selects a magnification factor according to the image resolution, and after determining the magnification factor, amplifies the image; a number is set for each area of ​​the image, and samples are taken in sequence according to the number and sent to the wavelet transformation unit.

[0043] In the intelligent image recognition and analysis system, the optimization of the feature extraction module is crucial to improving the recognition accuracy and efficiency. The feature extraction module in this embodiment includes an image magnification unit, a key area selection unit, a wavelet transformation unit and a key feature extraction module, which work together to achieve accurate extraction of image features.

[0044] In practical applications, such as product identification on an automated production line, the image acquisition module first acquires the entire image and numbers each area. Subsequently, the image magnification unit selects an appropriate magnification factor based on the image resolution and magnifies the image. This step ensures the clarity of image details in subsequent processing and improves the accuracy of feature extraction.

[0045] The wavelet transform unit separates the low-frequency and high-frequency signals of the enlarged image through convolution operations. The key region selection unit focuses on processing low-frequency signals and extracts feature vectors in each low-frequency region. These feature vectors are then sent to the key feature extraction module for further processing and analysis. This processing method not only reduces the complexity of data processing, but also improves the efficiency of feature extraction.

[0046] Compared with the prior art, the feature extraction module of this embodiment has the following outstanding creative features: Image magnification and preprocessing: By introducing an image magnification unit, this embodiment can adaptively magnify images of different resolutions, thereby ensuring the clarity of image details in subsequent processing. This preprocessing step is not common in the prior art and significantly improves the accuracy of feature extraction.

[0047] Key region selection and low-frequency feature extraction: The key region selection unit focuses on the processing of low-frequency signals, which helps to reduce the interference of noise and redundant information and improve the effectiveness of feature extraction. At the same time, low-frequency features often contain the main structural information of the image, which is of great significance for subsequent recognition and analysis.

[0048] Wavelet transform and multi-scale analysis: The wavelet transform unit separates low-frequency and high-frequency through convolution operation, which provides a basis for subsequent feature extraction. In addition, this implementation also introduces the concept of wavelet multi-scale decomposition, decomposing the image into low-frequency coefficients and high-frequency coefficients of different resolutions, which helps to capture the feature information of the image at different scales.

[0049] As a preferred implementation, when the image magnification unit selects a wavelet transform unit to sample each image magnification area after the area division, the sampling is performed by image stretching before and during the magnification sampling. The sampling steps are: firstly, edge analysis is performed on the image, and an edge extraction algorithm based on image grayscale difference is used to identify the edge of the image; the grayscale difference between the object in the image and the background feature image is set, and the grayscale difference is gradually increased from 0 to extract the target feature point. After the target feature point is successfully extracted, the target feature point is marked and the corresponding grayscale difference is recorded.

[0050] The feature extraction module marks the target feature points. If the grayscale difference is greater than 0 and less than the set a, a target feature point is marked every two points on the edge line, and a marked point is recorded between each two adjacent target feature points; if the grayscale difference is greater than the set a and less than the set b, each interval point on the edge line is marked as a target feature point, and a marked point is recorded between two adjacent target feature points; if the grayscale difference is greater than the set b and less than the set c, a target feature point is marked every 0.5 points on the edge line, and a marked point is recorded between two adjacent target feature points.

[0051] When the image magnification unit performs sampling, this embodiment also introduces image stretching technology. Before and during the magnification sampling, the edge of the image is identified and analyzed by an edge extraction algorithm based on image grayscale difference. This step helps to accurately extract target feature points and provides a basis for subsequent marking and processing.

[0052] In specific applications, such as the field of medical image analysis, image stretching technology can significantly improve the recognition accuracy of lesion areas in images. By performing edge analysis on the image and setting a suitable grayscale difference range, the target feature points of the lesion area can be accurately extracted. These feature points are then used for further diagnosis and analysis. Grayscale differential edge extraction in the image stretching and sampling technology of this embodiment: Through an edge extraction algorithm based on image grayscale difference, this embodiment can accurately identify edge information in the image and provide a basis for subsequent sampling and marking. This algorithm is not common in the prior art, and it significantly improves the extraction accuracy of target feature points. Adaptive sampling strategy: According to the set range of grayscale difference, this embodiment adopts an adaptive sampling strategy. When the grayscale difference is within a certain range, different sampling intervals and marking methods are used to ensure the accurate extraction and marking of target feature points.

[0053] The marking points are marked by setting colors so that pixels of corresponding colors are marked on the image: when the grayscale difference is greater than 0 and less than the set a, the marking points are displayed in red and the target feature points are displayed in blue; when the grayscale difference is greater than the set a and less than the set b, the marking points are displayed in white and the target feature points are displayed in pink; when the grayscale difference is greater than the set b and less than the set c, the marking points are displayed in black and the target feature points are displayed in light yellow. In the feature extraction process, accurate marking and color coding of target feature points are crucial for subsequent identification and analysis. In this embodiment, the target feature points and marking points are coded with different colors according to the set range of the grayscale difference.

[0054] In specific applications, such as vehicle recognition in intelligent transportation systems, by extracting and marking vehicle images, it is possible to quickly identify vehicle type, color and other information. Different color codes help to intuitively display the location of target feature points and marking points in the image, thereby improving the efficiency and accuracy of recognition.

[0055] The target feature point marking and color coding of this embodiment have the following characteristics: By encoding the target feature points and marking points with different colors, this embodiment intuitively displays the positions of these points in the image, which helps users quickly identify and understand the information in the image. The color coding strategy of this embodiment can be adjusted and expanded according to different application scenarios and requirements. For example, if higher recognition accuracy is required, more grayscale difference ranges and corresponding color codes can be added.

[0056] The feature extraction module performs wavelet multiscale decomposition on the image features: the wavelet multiscale decomposition of the image is decomposed into a low-frequency coefficient and several high-frequency coefficients of different resolutions, and then the eigenvalues ​​are extracted; the extracted eigenvalues ​​and the corresponding feature points when the eigenvalues ​​are extracted are subjected to feature conversion through the feature conversion formula, and the obtained feature point data are combined to obtain high-dimensional features. In the final stage of the feature extraction module, this embodiment introduces the concepts of wavelet multiscale decomposition and feature conversion. By decomposing the image into low-frequency coefficients and high-frequency coefficients of different resolutions, the feature information of the image at different scales can be captured. Subsequently, the extracted eigenvalues ​​and feature points are converted through the feature conversion formula to obtain high-dimensional feature data.

[0057] In specific applications, such as the field of remote sensing image processing, wavelet multiscale decomposition and feature conversion technology can significantly improve the recognition accuracy of ground objects in images. By performing multiscale decomposition and feature extraction on remote sensing images, accurate classification and recognition of different types of ground objects can be achieved. The wavelet multiscale decomposition and feature conversion of this embodiment have the following characteristics: Multiscale feature extraction: Through wavelet multiscale decomposition, this embodiment can capture the feature information of the image at different scales, providing richer feature data for subsequent recognition and analysis. Feature conversion and data combination: The extracted eigenvalues ​​and feature points are converted through the feature conversion formula, and the obtained feature point data is combined into high-dimensional feature data. This processing method not only improves the dimensionality and richness of the feature data, but also helps to improve the robustness and generalization ability of the recognition algorithm.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. Intelligent image recognition and analysis system, characterized in that: It includes data interaction processing among image acquisition module, feature extraction module, feature database module, comparison module and output module; The image acquisition module collects the image to be collected by partitioning, numbers each area and sends them to the feature extraction module in sequence according to the numbering sequence; The feature extraction module extracts high-dimensional features in each numbered image region by wavelet transform and key feature marking, and sends the extracted features to the feature database module; The feature database module is used to establish a feature database, perform fuzzy matching on the extracted high-dimensional features of the target image and the features in the feature database; establish a feature point control group through fuzzy matching, select approximate feature point control group data and send it to the comparison module; The comparison module performs corresponding analysis and comparison on the feature point control group matched by the feature extraction module and the feature database module, and after exporting the classification result of the target image, feeds it back to the output module; The output module receives the classification result of the target image, reorganizes and splices the image areas according to the numbers, and then displays them on the display screen.

2. The intelligent image recognition and analysis system according to claim 1, characterized in that: The feature extraction module includes an image magnification unit, a key area selection unit, a wavelet transformation unit and a key feature extraction module. When the image acquisition module extracts the entire image, the wavelet transformation unit separates the image from low frequency to high frequency through convolution; the key area selection unit acquires low-frequency signals, extracts feature vectors in each low-frequency area, and processes the feature vectors accordingly through the key feature extraction module; the image magnification unit selects a magnification factor according to the image resolution, and after determining the magnification factor, amplifies the image; a number is set for each area of ​​the image, and samples are taken in sequence according to the number and sent to the wavelet transformation unit.

3. The intelligent image recognition and analysis system according to claim 2, characterized in that: When the image magnification unit selects the wavelet transform unit to sample the image magnification area after each area division, the sampling is performed by image stretching before and during the magnification sampling. The sampling steps are: firstly, edge analysis is performed on the image, and an edge extraction algorithm based on image grayscale difference is adopted to identify the edge of the image; the grayscale difference between the object in the image and the background feature image is set, and the grayscale difference is gradually increased from 0 to extract the target feature point. After the target feature point is successfully extracted, the target feature point is marked and the corresponding grayscale difference is recorded.

4. The intelligent image recognition and analysis system according to claim 3, characterized in that: The feature extraction module marks the target feature points. If the grayscale difference is greater than 0 and less than the set a, a target feature point is marked every two points on the edge line, and a marked point is recorded between each two adjacent target feature points; if the grayscale difference is greater than the set a and less than the set b, each interval point on the edge line is marked as a target feature point, and a marked point is recorded between two adjacent target feature points; if the grayscale difference is greater than the set b and less than the set c, a target feature point is marked every 0.5 points on the edge line, and a marked point is recorded between two adjacent target feature points.

5. The intelligent image recognition and analysis system according to claim 4, characterized in that: The marking points are marked by setting colors so that pixels of corresponding colors are marked on the image: when the grayscale difference is greater than 0 and less than the set a, the marking points are displayed in red and the target feature points are displayed in blue; when the grayscale difference is greater than the set a and less than the set b, the marking points are displayed in white and the target feature points are displayed in pink; when the grayscale difference is greater than the set b and less than the set c, the marking points are displayed in black and the target feature points are displayed in light yellow.

6. The intelligent image recognition and analysis system according to claim 1, characterized in that: The feature extraction module performs wavelet multi-scale decomposition on the image features: decomposing the image into a low-frequency coefficient and a number of high-frequency coefficients with different resolutions, and then extracting feature values; The extracted eigenvalues ​​and the corresponding feature points when the eigenvalues ​​are extracted are transformed by the feature conversion formula, and the obtained feature point data are combined to obtain high-dimensional features.

Images