Space target image optimization and recognition system
By introducing image processing, feature analysis and target recognition modules into the spatial target image matching system, the recognition difficulty problems caused by noise and target diversity are solved, and more efficient and accurate spatial target image recognition is achieved.
Patent Information
- Application Number
- CN202510118127.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-24
- Publication Date
- 2025-05-27
AI Technical Summary
Existing spatial target image matching methods do not perform well in noisy, blurred or distorted image data, and the recognition difficulty increases due to the diversity of targets.
Provided is a spatial target image optimization and recognition system, including an image processing module, a feature analysis module and a target recognition module. The image processing module improves image quality and edge clarity through clear preprocessing and edge extraction, the feature analysis module improves recognition accuracy through feature spatial distribution optimization, and the target recognition module realizes real-time monitoring and tracking of new targets through target incremental recognition.
By optimizing image processing and feature analysis, the recognition accuracy and efficiency of spatial target images are improved, and more accurate identification and monitoring can be achieved in noise and diversity target environments.
Smart Images
Figure CN120047704A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and specifically to a spatial target image optimization and recognition system. Background Technique
[0002] In recent years, with the improvement of sensor technology, the data volume of spatial target images has also increased; how to use algorithms to extract the required target information from spatial target image data has become the focus of research. The spatial target image matching technology, which can obtain the spatial position relationship between images, is the basis for applications such as spatial target image stitching and three-dimensional reconstruction of spatial targets, and is of great significance.
[0003] Currently, most spatial target image matching methods directly use classical feature descriptors such as SIFT and ORB to extract feature points, then use feature distances for rough matching, and finally use RANSAC to eliminate mis-matched points to obtain the final matching point pairs. Due to the following characteristics of the spatial target structure itself: one is the data quality problem: the image data may have problems such as noise, blur, or distortion, which affect the performance of the system. The other is the target diversity: the shapes, sizes, and features of spatial targets may be very diverse, resulting in increased recognition difficulty. Summary of the Invention
[0004] The purpose of the present invention is to provide a spatial target image optimization and recognition system to solve the problems raised in the above background technique.
[0005] To achieve the above purpose, the present invention provides the following technical solution: A spatial target image optimization and recognition system includes an image processing module, a feature analysis module, and a target recognition module. The image processing module includes image clarity preprocessing and edge extraction, the feature analysis module includes feature spatial distribution optimization, and the target recognition module includes target incremental recognition.
[0006] The image clarity preprocessing is used to improve the quality and clarity of the image.
[0007] The edge extraction is used to accurately extract the target edges in the image, providing a basis for subsequent processing.
[0008] The feature spatial distribution optimization is used to analyze and optimize the features of the target, improving the accuracy and efficiency of recognition.
[0009] The target incremental recognition is used to be able to recognize newly emerging targets, realizing real-time monitoring and tracking of spatial targets.
[0010] Preferably, the image clarity preprocessing includes denoising, contrast enhancement, image sharpening, color correction, geometric correction, and image segmentation.
[0011] The denoising removes the noise in the image and improves the image quality. The denoising methods include mean filtering, median filtering or Gaussian filtering.
[0012] The contrast enhancement makes the difference between the target and the background more obvious by adjusting the brightness and contrast of the image, which is helpful for subsequent edge extraction and target recognition.
[0013] The image sharpening enhances the edges and details of the image, making the contour of the target clearer.
[0014] The color correction performs operations such as color space conversion and white balance correction on the image to make it more in line with the human eye observation habit.
[0015] The geometric correction performs operations such as distortion and rotation on the image to correct the geometric deformation of the image.
[0016] The image segmentation divides the image into different regions or objects to facilitate the separate processing and analysis of the target.
[0017] Preferably, in the present invention, the edge extraction adopts the gray difference method, and determines the edges of the image by calculating the gray difference between adjacent pixels.
[0018] The edge extraction also includes determining the neighborhood size used when calculating the gray difference. A smaller neighborhood can detect more detailed edges, while a larger neighborhood can smooth the noise and detect more global edges; selecting an appropriate threshold to determine which gray differences are considered as edges. The selection of the threshold will affect the result of edge detection and can be adjusted through experiments or according to the characteristics of the image.
[0019] Preferably, in the present invention, before applying the gray difference method, preprocessing of smoothing filtering is performed on the image to reduce the influence of noise on edge detection; post-processing is performed on the detected edges, and the post-processing includes morphological operations such as dilation and erosion to further optimize the continuity and accuracy of the edges.
[0020] Preferably, in the present invention, the optimization of the feature space distribution includes feature selection, feature extraction, feature transformation, feature combination and feature optimization.
[0021] The feature selection selects the features that are most helpful for target recognition from the original features, reduces the feature dimension, and improves the recognition efficiency and accuracy.
[0022] The feature extraction converts the original image into more representative and discriminative features through analysis methods, and the analysis methods include principal component analysis (PCA) and linear discriminant analysis (LDA).
[0023] The feature transformation transforms the extracted features to make them more suitable for classification or recognition tasks, and uses a kernel function to map the features to a high-dimensional space.
[0024] The feature combination combines the shape features and texture features to obtain a more comprehensive and accurate description of the target.
[0025] The feature optimization uses a genetic algorithm or a particle swarm optimization algorithm to adjust and optimize the features, improving the quality and distinctiveness of the features.
[0026] Preferably, the target incremental recognition of the present invention includes new target detection, target feature extraction, feature matching and classification, model update, and incremental learning.
[0027] The new target detection detects new targets that appear in the image through the analysis and processing of the image.
[0028] The target feature extraction extracts the features of the new target for comparison and recognition with the existing target features.
[0029] The feature matching and classification matches and classifies the features of the new target with the existing target features to determine the category of the new target.
[0030] The model update updates the target recognition model according to the features and classification results of the new target to improve the model's recognition ability for new targets.
[0031] The incremental learning uses the method of incremental learning to continuously add the features and categories of new targets to the model, realizing the continuous learning and optimization of the model.
[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: Through the image processing module, the present invention can improve the clarity, contrast, and resolution of the space target image, making the target more clearly distinguishable. The optimized image can provide more details and features, which helps to improve the accuracy and reliability of the recognition system, provides more accurate data for space science research, and helps scientists better understand the characteristics and behaviors of space targets. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 It is a schematic diagram of the system of the present invention. DETAILED DESCRIPTION OF THE INVENTION
[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0035] The present invention provides a technical solution: a space target image optimization and recognition system, including an image processing module, a feature analysis module, and a target recognition module.
[0036] Embodiment 1 Please refer to Figure 1 , this embodiment provides an image processing module in a space target image optimization and recognition system. The image processing module includes clarity preprocessing and edge extraction.
[0037] The clarity preprocessing is used to improve the quality and clarity of the image.
[0038] The edge extraction is used to accurately extract the target edge in the image, providing a basis for subsequent processing.
[0039] The clarity preprocessing includes denoising, contrast enhancement, image sharpening, color correction, geometric correction, and image segmentation.
[0040] The denoising removes the noise in the image and improves the image quality. The denoising methods include mean filtering, median filtering, or Gaussian filtering.
[0041] The contrast enhancement makes the difference between the target and the background more obvious by adjusting the brightness and contrast of the image, which helps with subsequent edge extraction and target recognition.
[0042] The image sharpening enhances the edges and details of the image, making the contour of the target clearer.
[0043] The color correction performs operations such as color space conversion and white balance correction on the image to make it more in line with the human eye's viewing habits.
[0044] The geometric correction performs operations such as distortion and rotation on the image to correct the geometric deformation of the image.
[0045] The image segmentation divides the image into different regions or objects to facilitate the separate processing and analysis of the target.
[0046] These image clarity preprocessing methods can be selected and combined according to specific application requirements and image characteristics to improve the performance and accuracy of the space target image optimization and recognition system.
[0047] The edge extraction adopts the gray difference method, which determines the edges of an image by calculating the gray differences between adjacent pixels. The Canny operator, Sobel operator or Laplacian operator can also be used.
[0048] Edge extraction also includes determining the neighborhood size used when calculating the gray differences. A smaller neighborhood can detect more detailed edges, while a larger neighborhood can smooth the noise and detect more global edges; selecting an appropriate threshold to determine which gray differences are considered as edges. The selection of the threshold will affect the result of edge detection and can be adjusted through experiments or according to the characteristics of the image.
[0049] Before applying the gray difference method, preprocessing of smoothing filtering is performed on the image to reduce the influence of noise on edge detection; post-processing is performed on the detected edges, and the post-processing includes morphological operations such as dilation and erosion to further optimize the continuity and accuracy of the edges.
[0050] The gray difference method is a method for determining the edges of an image by calculating the gray differences between adjacent pixels. The following are some refined aspects of the gray difference method.
[0051] The gray difference method can use different calculation methods, such as horizontal difference, vertical difference, diagonal difference, etc. Different calculation methods may be sensitive to different types of edges. In practical applications, usually an appropriate edge extraction method is selected according to specific situations, or multiple methods are combined for edge extraction to improve the accuracy and reliability of edge detection.
[0052] Embodiment 2 Please refer to Figure 1 , this embodiment provides a feature analysis module in a spatial target image optimization and recognition system, and the feature analysis module includes feature space distribution optimization.
[0053] The feature space distribution optimization includes feature selection, feature extraction, feature transformation, feature combination and feature optimization.
[0054] The feature space distribution optimization includes feature selection, feature extraction, feature transformation, feature combination and feature optimization.
[0055] The feature selection selects the most helpful features for target recognition from the original features, reduces the feature dimension, and improves the recognition efficiency and accuracy.
[0056] The feature extraction converts the original image into more representative and discriminative features through analysis methods, and the analysis methods include principal component analysis (PCA), linear discriminant analysis (LDA).
[0057] The feature transformation transforms the extracted features to make them more suitable for classification or recognition tasks, and uses a kernel function to map the features to a high-dimensional space.
[0058] The feature combination combines shape features and texture features to obtain a more comprehensive and accurate description of the target.
[0059] The feature optimization uses a genetic algorithm or a particle swarm optimization algorithm to adjust and optimize the features, improving the quality and distinctiveness of the features.
[0060] Using linear discriminant, calculate the between-class scatter matrix and the within-class scatter matrix. Solve for the projection direction that maximizes the ratio of the between-class scatter to the within-class scatter. The resulting projection direction is the result of linear discriminant analysis. The advantages of linear discriminant analysis include: simple and intuitive, easy to understand and implement. It can effectively reduce the data dimension. It has no specific requirements for the data distribution.
[0061] The basic idea of using principal component analysis is to combine the original variables into a new set of variables through linear transformation, so that these new variables are uncorrelated with each other and can retain as much information of the original variables as possible. By converting multiple correlated variables into a few uncorrelated principal components, the data dimension can be reduced and the data complexity can be lowered. The principal components can retain as much information of the original variables as possible, so the main information in the data can be extracted. Principal component analysis can convert high-dimensional data into low-dimensional data, making the data easier to visualize.
[0062] The application fields of principal component analysis are very extensive, including: data analysis, pattern recognition, image processing, etc.
[0063] Embodiment 3 Please refer to Figure 1 , this embodiment provides a target recognition module in a space target image optimization and recognition system. The target recognition module includes target incremental recognition.
[0064] The target incremental recognition includes new target detection, target feature extraction, feature matching and classification, model update, and incremental learning.
[0065] The new target detection detects new targets that appear in the image through the analysis and processing of the image.
[0066] The target feature extraction extracts the features of the new target for comparison and recognition with the existing target features.
[0067] The feature matching and classification matches and classifies the features of the new target with the existing target features to determine the category of the new target.
[0068] The model is updated. Based on the features and classification results of new targets, the target recognition model is updated to improve the model's recognition ability for new targets.
[0069] For the incremental learning, an incremental learning method is adopted to continuously add the features and categories of new targets into the model, realizing the continuous learning and optimization of the model.
[0070] It can improve the recognition ability and accuracy of the space target image optimization and recognition system for new targets, enabling it to better adapt to the changing space environment and mission requirements.
[0071] Based on what is described in Embodiments 1 - 3, the space target image optimization and recognition system aims to enhance the observation, recognition, and analysis capabilities of space targets. Through advanced image processing techniques and machine learning algorithms, the system can achieve precise positioning, feature extraction, and intelligent recognition of space targets.
[0072] The present invention plays a fundamental and crucial role in military operations, and can perform positioning, regional super - resolution reconstruction, and type recognition of space targets such as satellites, space debris, and space vehicles.
[0073] It is mainly applied to the aerospace field, including space target surveillance, recognition, and tracking, etc.
[0074] Through this system, the optimization and recognition of space target images are jointly achieved.
[0075] It should be noted that: the entire device is controlled through a master control button. Since the device matched with the control button is a common device and belongs to the existing mature technology, the electrical connection relationship and the specific circuit structure are not elaborated herein.
[0076] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A space target image optimization and recognition system, characterized in that: It includes an image processing module, a feature analysis module, and a target recognition module. The image processing module includes clearing preprocessing and edge extraction, the feature analysis module includes feature space distribution optimization, and the target recognition module includes target increment recognition; The clarity preprocessing is used to improve the quality and clarity of the image; The edge extraction is used to accurately extract the target edge in the image, providing a basis for subsequent processing; The feature space distribution optimization is used to analyze and optimize the features of the target to improve the accuracy and efficiency of recognition; The target incremental recognition is used to identify newly appeared targets and realize real-time monitoring and tracking of space targets.
2. The space target image optimization and recognition system according to claim 1, characterized in that: The clarity preprocessing includes denoising, contrast enhancement, image sharpening, color correction, geometric correction and image segmentation; The denoising method removes noise in the image to improve the image quality. The denoising method includes mean filtering, median filtering or Gaussian filtering; The contrast enhancement adjusts the brightness and contrast of the image to make the difference between the target and the background more obvious, which is helpful for subsequent edge extraction and target recognition; The image sharpening enhances the edges and details of the image and makes the outline of the target clearer; The color correction is to convert the color space and correct the white balance of the image to make it more in line with human eye observation habits; The geometric correction is to distort and rotate the image to correct the geometric deformation of the image; The image segmentation divides the image into different regions or objects so as to facilitate separate processing and analysis of the targets.
3. The space target image optimization and recognition system according to claim 1, characterized in that: The edge extraction adopts the grayscale difference method to determine the edge of the image by calculating the grayscale difference between adjacent pixels; Edge extraction also includes determining the neighborhood size used when calculating grayscale differences. Smaller neighborhoods can detect more detailed edges, while larger neighborhoods can smooth noise and detect more global edges. Selecting an appropriate threshold determines which grayscale differences are considered edges. The choice of threshold will affect the results of edge detection and can be adjusted through experiments or according to image characteristics.
4. The space target image optimization and recognition system according to claim 3, characterized in that: Before the grayscale difference method is applied, the image is pre-processed by smoothing filtering to reduce the influence of noise on edge detection; the detected edge is post-processed, and the post-processing includes morphological operations of expansion and erosion to further optimize the continuity and accuracy of the edge.
5. The space target image optimization and recognition system according to claim 1, characterized in that: The feature space distribution optimization includes feature selection, feature extraction, feature transformation, feature combination and feature optimization; The feature selection selects the features that are most helpful for target recognition from the original features, reduces the feature dimension, and improves recognition efficiency and accuracy; The feature extraction converts the original image into more representative and distinguishing features through an analysis method, wherein the analysis method includes principal component analysis and linear discriminant analysis; The feature transformation transforms the extracted features to make them more suitable for classification or recognition tasks, and uses a kernel function to map the features to a high-dimensional space; The feature combination combines shape features and texture features to obtain a more comprehensive and accurate target description; The feature optimization uses a genetic algorithm or a particle swarm optimization algorithm to adjust and optimize the features to improve the quality and distinguishability of the features.
6. The space target image optimization and recognition system according to claim 1, characterized in that: The target incremental recognition includes new target detection, target feature extraction, feature matching and classification, model updating and incremental learning; The new target detection is to detect new targets appearing in the image by analyzing and processing the image; The target feature extraction is to extract the features of the new target so as to compare and identify them with the existing target features; The feature matching and classification is to match and classify the features of the new target with the features of the existing targets to determine the category of the new target; The model updating is to update the target recognition model according to the characteristics and classification results of the new target to improve the model's recognition ability for the new target; The incremental learning adopts an incremental learning method to continuously add features and categories of new targets into the model to achieve continuous learning and optimization of the model.