Intelligent splicing method and system for electronic images based on decision network

Abstract

The present invention relates to the field of intelligent image splicing, and proposes an intelligent splicing method and system for electronic images based on a decision network. The method comprises: extracting feature descriptors of images to be spliced ​​using a preset convolutional neural network, and calculating feature similarity values, performing image registration using the feature similarity values, performing region segmentation using a preset clustering algorithm, extracting key feature points and calculating texture difference items, generating a splicing candidate set based on the texture difference items, and performing error detection using an automatic detection mechanism, inputting error detection records into an image splicing processing system for correction, extracting correction factors to generate a corrected image set, removing the splicing edges of the corrected image set, performing color balancing, and optimizing the splicing of images using a particle swarm optimization algorithm, and finally obtaining an optimized spliced ​​image. The present invention can improve the splicing accuracy and splicing efficiency of electronic images.

Description

Technical Field

[0001] The present invention relates to the field of intelligent image splicing, and in particular to a method and system for realizing intelligent splicing of electronic images based on a decision network. Background Art

[0002] Electronic images are images captured by optical, electronic or digital devices, which can be processed and stored on computers or other digital devices. They are widely used in medical diagnosis, remote sensing, security monitoring and other fields.

[0003] At present, there are two main image stitching methods for electronic images, namely, those based on feature point matching and those based on optimization algorithms. The method based on feature point matching realizes image stitching by extracting feature points in the image and matching the positional relationship of these feature points. When there are occlusions, lighting changes, or geometric distortions in the image, feature point matching is prone to errors, resulting in inaccurate stitching results. The method based on optimization algorithms optimizes the energy function of the stitching results to realize image stitching, which requires multiple iterative calculations, and each iteration requires calculating the value of the energy function and updating the variables, which may lead to high computational complexity and affect the image stitching efficiency. Therefore, an intelligent stitching method for electronic images is needed to improve the stitching accuracy and efficiency of electronic images. Summary of the invention

[0004] The present invention provides a method and system for realizing intelligent splicing of electronic images based on a decision network, the main purpose of which is to improve the splicing accuracy and splicing efficiency of electronic images.

[0005] Acquire images to be stitched, extract feature descriptors corresponding to the images to be stitched using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be stitched based on the feature descriptors;

[0006] Based on the feature similarity values, the images to be stitched are registered to obtain a registered image set, the registered image set is segmented by a preset clustering algorithm to obtain an image segmentation region, key feature points in the image segmentation region are extracted, and the saturation and color difference corresponding to the key feature points are queried, and based on the saturation and the color difference, a texture difference item corresponding to the image segmentation region is calculated;

[0007] Based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images is generated, and an error detection is performed on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record;

[0008] After the error detection record is input into a preset image stitching processing system, the error detection record is corrected using the fault tolerance mechanism in the image stitching processing system to obtain a corrected data set, a corresponding correction factor in the corrected data set is extracted, and a corrected image set corresponding to the stitching candidate set is generated based on the correction factor;

[0009] The stitching edges of the images in the corrected image set are removed to obtain an edge-removed image set, the edge-removed image set is color balanced to obtain a balanced image set, and the images in the balanced image set are optimized and stitched using a preset particle swarm optimization algorithm to obtain optimized stitched images corresponding to the images to be stitched.

[0010] Optionally, the calculating, based on the feature descriptors, feature similarity values ​​corresponding to the images to be stitched includes:

[0011] Identify a descriptor dimension corresponding to the feature descriptor, and query an element scale corresponding to an element in the descriptor dimension;

[0012] Based on the descriptor dimension and the element scale, the feature similarity value corresponding to the image to be stitched is calculated using the following formula:

[0013]

[0014] in, represents the feature similarity value corresponding to the image to be stitched, refers to the descriptor dimension, represents the element index in the descriptor dimension, Indicates the descriptor Middle element values, Indicates the descriptor Middle element values, refers to the element size.

[0015] Optionally, the performing image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set includes:

[0016] Based on the feature similarity value, query the spatial position corresponding to the feature element in the image to be stitched;

[0017] Based on the spatial position, calculating the transformation matrix corresponding to the characteristic elements in the image to be stitched;

[0018] Using the transformation matrix, the characteristic elements in the images to be spliced ​​are mapped to the same coordinates to obtain a characteristic element coordinate system;

[0019] Based on the characteristic element coordinate system, the images to be spliced ​​are registered to obtain a registered image set.

[0020] Optionally, the using a preset clustering algorithm to perform region segmentation on the registered image set to obtain image segmentation regions includes:

[0021] Preprocessing the images in the registration image set to obtain preprocessed images;

[0022] After extracting the image pixel values ​​in the preprocessed image, calculating the statistical features corresponding to the image pixel values;

[0023] Randomly selecting k features from the statistical features as initial cluster centers, assigning each pixel in the preprocessed image to the nearest initial cluster center, and calculating the average pixel value in the initial cluster center;

[0024] Based on the pixel average value, the registered image set is segmented to obtain image segmentation regions.

[0025] Optionally, the calculating, based on the saturation and the color difference, a texture difference item corresponding to the image segmentation area includes:

[0026] Based on the saturation and the color difference, querying the regional feature points in the image segmentation area;

[0027] Based on the regional feature points, the texture difference item corresponding to the image segmentation area is calculated using the following formula:

[0028]

[0029] in, represents the texture difference item corresponding to the image segmentation area, represents the index corresponding to the regional feature point, n represents the total number of regional feature points, It represents the saturation value of the feature point in the jth region. represents the weight of the j-th regional feature point, Represents the color difference value of the j-th region feature point.

[0030] Optionally, the generating, based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images includes:

[0031] Calculating a similarity measure corresponding to the texture difference item;

[0032] Based on the similarity measure, query the image sorting area corresponding to the texture difference item;

[0033] Selecting the image with the highest similarity in the image sorting area as a candidate image;

[0034] The candidate images in the image selection area are combined to obtain a splicing candidate set corresponding to the image to be spliced.

[0035] Optionally, generating a corrected image set corresponding to the stitching candidate set based on the correction factor includes:

[0036] Querying the factor data sequence corresponding to the correction factor;

[0037] Performing sequence analysis on the factor data sequence to obtain a plurality of correction parameters, the plurality of correction parameters including: a color shift parameter, a distortion correction parameter, and a brightness adjustment parameter;

[0038] Based on the color shift parameter, the distortion correction parameter, and the brightness adjustment parameter, image correction optimization is performed on the stitching candidate set to obtain a corrected image set corresponding to the stitching candidate set.

[0039] Optionally, removing the spliced ​​edges of the images in the corrected image set to obtain the edge-removed image set includes:

[0040] Performing gradient calculation on the images in the correction image set to obtain a gradient image;

[0041] Performing threshold segmentation on the gradient image to obtain a binary image;

[0042] Performing morphological connection on the binary images to obtain a connected morphological image;

[0043] extracting an edge mask in the connected morphological image;

[0044] The edge mask is used to remove the spliced ​​edges of the images in the corrected image set to obtain an edge-removed image set.

[0045] Optionally, performing color balancing on the edge-removed image set to obtain a balanced image set includes:

[0046] Traversing the pixel points of the image in the edge-removed image set, and calculating the average color value corresponding to the pixel points;

[0047] Based on the average color value, adjusting the contrast of the images in the edge-removed image set;

[0048] Based on the contrast, matching the color saturation of the images in the edge-removed image set;

[0049] Based on the color saturation, color balancing is performed on the edge-removed image set to obtain a balanced image set.

[0050] In order to solve the above problems, the present invention also provides an intelligent splicing system for electronic images based on a decision network, the system comprising:

[0051] A feature similarity value calculation module is used to obtain the images to be spliced, extract feature descriptors corresponding to the images to be spliced ​​using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be spliced ​​based on the feature descriptors;

[0052] a texture difference item calculation module, configured to perform image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, perform region segmentation on the registered image set using a preset clustering algorithm to obtain image segmentation regions, extract key feature points in the image segmentation regions, query the saturation and color difference corresponding to the key feature points, and calculate the texture difference item corresponding to the image segmentation regions based on the saturation and the color difference;

[0053] An error detection module, used to generate a stitching candidate set corresponding to the to-be-stitched images based on the texture difference item, and perform error detection on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record;

[0054] An image correction module is used to input the error detection record into a preset image stitching processing system, and then use the fault tolerance mechanism in the image stitching processing system to correct the error detection record to obtain a corrected data set, extract the corresponding correction factor in the corrected data set, and generate a corrected image set corresponding to the stitching candidate set based on the correction factor;

[0055] The optimization stitching module is used to remove the stitching edges of the images in the corrected image set to obtain an edge-removed image set, perform color balance on the edge-removed image set to obtain a balanced image set, and use a preset particle swarm optimization algorithm to optimize the stitching of the images in the balanced image set to obtain optimized stitching images corresponding to the images to be stitched.

[0056] First, the present invention obtains the images to be stitched, and uses a preset convolutional neural network to extract the feature descriptors corresponding to the images to be stitched, so as to learn the local and global features of the images, and the feature descriptors can more accurately represent the image content and enhance the features of the images, thereby improving the accuracy and stability of image stitching. At the same time, the present invention performs image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, thereby improving the quality and accuracy of image stitching, and can better preserve the information and features of the original images, while providing multi-view and consistent observation. The present invention generates a stitching candidate set corresponding to the images to be stitched based on the texture difference items, which can improve the stitching effect and improve automatic The present invention improves the degree of digitization, increases flexibility and reduces the stitching error rate, which has important benefits for image stitching tasks. Secondly, the present invention inputs the error detection record into a preset image stitching processing system, and then uses the fault-tolerant mechanism in the image stitching processing system to correct the error detection record to obtain a corrected data set, which can improve the accuracy, efficiency and robustness of the stitching result, while reducing the need for subsequent intervention, and can more robustly handle various stitching scenarios and improve overall stability. Furthermore, the present invention removes the stitching edges of the images in the corrected image set to obtain an edge-removed image set, which can improve image quality, improve image analysis results, simplify subsequent processing procedures, enhance subsequent processing experience, and reduce data storage and transmission costs. Therefore, the intelligent stitching method for electronic images based on a decision network proposed in the present invention can improve the stitching accuracy and efficiency of electronic images. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a flow chart of a method for intelligently stitching electronic images based on a decision network provided by an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of a module of a system for intelligently stitching electronic images based on a decision network provided by an embodiment of the present invention;

[0059] Figure 3 A schematic diagram of the internal structure of an electronic device for implementing an intelligent splicing method of electronic images based on a decision network provided by an embodiment of the present invention.

[0060] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0062] The embodiment of the present application provides a method for intelligent splicing of electronic images based on a decision network. The execution subject of the method for intelligent splicing of electronic images based on a decision network includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for intelligent splicing of electronic images based on a decision network can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0063] Reference Figure 1 FIG. 1 is a flow chart of a method for realizing intelligent splicing of electronic images based on a decision network according to an embodiment of the present invention. In this embodiment, the method for realizing intelligent splicing of electronic images based on a decision network includes:

[0064] S1. Acquire images to be stitched, use a preset convolutional neural network to extract feature descriptors corresponding to the images to be stitched, and calculate feature similarity values ​​corresponding to the images to be stitched based on the feature descriptors.

[0065] The present invention acquires images to be stitched and uses a preset convolutional neural network to extract feature descriptors corresponding to the images to be stitched, so as to learn local and global features of the images. The feature descriptors can more accurately represent the image content and enhance the features of the images, thereby improving the accuracy and stability of image stitching.

[0066] Among them, the preset convolutional neural network refers to a pre-trained model that has been trained on a large-scale data set and has good performance, such as VGG, ResNet, Inception and other models; the preset convolutional neural network is generally composed of a convolutional layer, a pooling layer, an activation function and a fully connected layer; the feature descriptor refers to a high-dimensional feature vector or feature map extracted from the convolutional neural network, and the output of the last fully connected layer is usually used as the feature descriptor, and these descriptors can contain local and global features of the image. Optionally, the extraction of the feature descriptors corresponding to the images to be spliced ​​can be obtained by utilizing a preset convolutional neural network.

[0067] Further, as an embodiment of the present invention, the calculating the feature similarity value corresponding to the image to be stitched based on the feature descriptor includes: identifying the descriptor dimension corresponding to the feature descriptor; querying the element scale corresponding to the element in the descriptor dimension; and calculating the feature similarity value corresponding to the image to be stitched based on the descriptor dimension and the element scale using the following formula:

[0068]

[0069] in, represents the feature similarity value corresponding to the image to be stitched, refers to the descriptor dimension, represents the element index in the descriptor dimension, Indicates the descriptor Middle element values, Indicates the descriptor Middle element values, refers to the element size.

[0070] It should be noted that, in the present invention, the descriptor dimension refers to the number of elements contained in the feature descriptor; the element scale refers to the ratio or scale of each element in the descriptor; and the feature descriptors corresponding to the images to be stitched are divided into two groups of descriptors, namely and .

[0071] S2. Based on the feature similarity values, the images to be stitched are registered to obtain a registered image set, the registered image set is segmented by a preset clustering algorithm to obtain image segmentation regions, key feature points in the image segmentation regions are extracted, and the saturation and color difference corresponding to the key feature points are queried, and the texture difference item corresponding to the image segmentation region is calculated based on the saturation and the color difference;

[0072] The present invention performs image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, thereby improving the quality and accuracy of image stitching, and can better preserve the information and features of the original images while providing multi-viewing angles and consistent observation.

[0073] The registered image set refers to a set of aligned images obtained after feature matching, transformation matrix calculation and image registration operations.

[0074] As an embodiment of the present invention, the image registration of the images to be spliced ​​based on the feature similarity value to obtain a registered image set includes: querying the spatial position corresponding to the feature element in the image to be spliced ​​based on the feature similarity value; calculating the transformation matrix corresponding to the feature element in the image to be spliced ​​based on the spatial position; using the transformation matrix, mapping the feature elements in the image to be spliced ​​to the same coordinate to obtain a feature element coordinate system; based on the feature element coordinate system, performing image registration on the images to be spliced ​​to obtain a registered image set.

[0075] Among them, the spatial position refers to the exact position coordinates of each feature element in the image to be stitched, usually represented by the pixel coordinates of the feature point; the transformation matrix refers to the matrix that describes the spatial relationship between the feature elements; the feature element coordinate system refers to the transformation matrix obtained by applying calculation, which maps the feature elements in the image to be stitched into a common coordinate system so that they are aligned in space, thereby facilitating subsequent image registration operations.

[0076] Furthermore, the querying of the spatial position corresponding to the characteristic elements in the images to be stitched can be obtained by a feature detection and matching algorithm, such as SIFT, SURF or ORB algorithms; the calculation of the transformation matrix corresponding to the characteristic elements in the images to be stitched can be obtained by a camera geometry model based on epipolar geometry, such as an affine transformation matrix or a projection transformation matrix; the mapping of the characteristic elements in the images to be stitched to the same coordinates can be obtained by utilizing the transformation matrix; the image registration of the images to be stitched can be obtained by a registration algorithm, such as an algorithm based on feature point matching, brightness or gradient information, and the like.

[0077] The present invention uses a preset clustering algorithm to perform region segmentation on the registered image set to obtain image segmentation regions, which can help extract feature information of each region, and then be used for subsequent image analysis, target recognition or scene understanding tasks, which helps to better understand the image content.

[0078] The image segmentation region refers to dividing the image into different regions according to the clustering result.

[0079] As an embodiment of the present invention, the method of performing regional segmentation on the registered image set using a preset clustering algorithm to obtain image segmentation regions includes: preprocessing the images in the registered image set to obtain preprocessed images; extracting image pixel values ​​in the preprocessed images, and calculating statistical features corresponding to the image pixel values; randomly selecting k features from the statistical features as initial clustering centers, and assigning each pixel in the preprocessed image to the nearest initial clustering center, and calculating the pixel average value in the initial clustering center; based on the pixel average value, performing regional segmentation on the registered image set to obtain image segmentation regions.

[0080] Among them, the preprocessed image refers to the image obtained after the preprocessing operation is performed on the registered image set, and the preprocessing operation may include denoising, smoothing, contrast enhancement and other operations; the image pixel value refers to the value of each pixel in the image, usually expressed as a grayscale value (in a grayscale image) or an RGB value (in a color image); the statistical feature refers to the image feature calculated based on the image pixel value, which may include mean, variance, energy, entropy, etc.; the initial cluster center refers to the k feature values ​​randomly selected at the beginning by the preset clustering algorithm, where K is the pre-set number of clusters; the pixel average value refers to the average value of the pixel values ​​contained in each cluster center.

[0081] Furthermore, the preprocessing of the images in the registered image set can be achieved through image processing library tools, such as OpenCV, Python's PIL library and other tools; the extraction of image pixel values ​​in the preprocessed images can be achieved through the NumPy library in Python; the calculation of statistical features corresponding to the image pixel values ​​can be achieved through a texture analysis model, such as GLCM, LBP, etc.; the random selection of k features as initial clustering centers among the statistical features and the assignment of each pixel in the preprocessed image to the nearest initial clustering center can be achieved through a preset clustering algorithm; the regional segmentation of the registered image set can be achieved through a clustering-based image segmentation method, such as K-means algorithm, MeanShift algorithm, etc.

[0082] The present invention can help identify and describe the most representative and important position or structural information in the image by extracting key feature points in the image segmentation area and querying the saturation and color difference corresponding to the key feature points, which is helpful for subsequent analysis and processing.

[0083] Among them, the key feature points refer to points or structures with significant properties, uniqueness and importance in the image segmentation area; the saturation refers to the purity or vividness of the color; the color difference refers to the degree of color change between different areas or pixels. Optionally, the extraction of the key feature points in the image segmentation area can be obtained by a feature extraction method, such as: Harris corner detection algorithm, Shi-Tomasi corner detection algorithm, etc.; the query of the saturation and color difference corresponding to the key feature points can be obtained by OpenCV or PIL library in Python.

[0084] Further, as an embodiment of the present invention, the calculating the texture difference item corresponding to the image segmentation area based on the saturation and the color difference includes: querying the regional feature points in the image segmentation area based on the saturation and the color difference; and calculating the texture difference item corresponding to the image segmentation area based on the regional feature points using the following formula:

[0085]

[0086] in, represents the texture difference item corresponding to the image segmentation area, represents the index corresponding to the regional feature point, n represents the total number of regional feature points, It represents the saturation value of the feature point in the jth region. represents the weight of the j-th regional feature point, Represents the color difference value of the j-th region feature point.

[0087] It should be noted that, in the present invention, the regional feature point refers to a specific pixel point in the image segmentation area; the texture difference item refers to a comprehensive value representing the texture difference between different regional feature points; the color difference value refers to the numerical difference between specific regional feature points in the image segmentation area in the color space (such as RGB, CMYK, HSV, etc.).

[0088] S3. Based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images is generated, and an error detection is performed on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record.

[0089] The present invention generates a stitching candidate set corresponding to the to-be-stitched images based on the texture difference items, which can improve the stitching effect, increase the degree of automation, increase flexibility and reduce the stitching error rate, and has important benefits for image stitching tasks.

[0090] The splicing candidate set refers to selecting an image that is most similar to the texture difference item of the region in each image sorting region as a candidate image.

[0091] As an embodiment of the present invention, generating a stitching candidate set corresponding to the image to be stitched based on the texture difference item includes: calculating a similarity measure corresponding to the texture difference item; querying an image sorting area corresponding to the texture difference item based on the similarity measure; selecting an image with the highest similarity in the image sorting area as a candidate image; and combining the candidate images in the image sorting area to obtain a stitching candidate set corresponding to the image to be stitched.

[0092] Among them, the similarity metric refers to a metric method used to measure the texture differences between different regions, and is used to calculate the similarity between image regions; the image sorting region refers to dividing the image to be spliced ​​into multiple local regions, each region represents an image segmentation region; the candidate image refers to selecting an image in each image sorting region that is most similar to the texture difference item of the region as a candidate image.

[0093] Furthermore, the calculation of the similarity measure corresponding to the texture difference item can be obtained through a feature extraction algorithm, such as: local binary pattern, directional gradient histogram and other algorithms; the query of the image sorting area corresponding to the texture difference item can be obtained through a superpixel-based segmentation method, such as: SLIC algorithm, QuickShift algorithm and the like; the selection of the image with the highest similarity in the image sorting area as the candidate image can be obtained through a feature matching algorithm, such as: SIFT, SURF, ORB and other feature descriptor algorithms; the image combination of the candidate images in the image sorting area can be obtained through an image stitching algorithm, such as: a panoramic stitching algorithm based on feature point matching, a stitching algorithm based on graph cuts and other algorithms.

[0094] The present invention uses a preset automatic detection mechanism to perform error detection on the splicing candidate set to obtain an error detection record, which can improve the accuracy, efficiency and robustness of the splicing result, reduce the need for subsequent intervention, and bring many benefits to the splicing task.

[0095] Among them, the preset automatic detection mechanism refers to a pre-set mechanism for error detection of the splicing candidate set. This mechanism can determine whether there are errors or mismatches in the candidate image by analyzing the characteristics of the image, such as boundary matching, color consistency, geometric shape, etc., according to specific problem areas and needs; the error detection record refers to the process and results of recording the error information found during the automatic error detection process. The error detection record includes the type, location, description and other relevant information of the error. Optionally, the error detection of the splicing candidate set can be achieved by utilizing the preset automatic detection mechanism.

[0096] S4. After inputting the error detection record into a preset image stitching processing system, the error detection record is corrected using the fault-tolerant mechanism in the image stitching processing system to obtain a corrected data set, and the corresponding correction factors in the corrected data set are extracted. Based on the correction factors, a corrected image set corresponding to the stitching candidate set is generated.

[0097] The present invention inputs the error detection record into a preset image stitching processing system, and then uses the fault-tolerant mechanism in the image stitching processing system to correct the error detection record to obtain a corrected data set, thereby improving the accuracy, efficiency and robustness of the stitching result, while reducing the need for subsequent intervention, and can more robustly handle various stitching scenarios, thereby improving overall stability.

[0098] Among them, the preset image stitching processing system refers to a pre-set hardware system for image stitching, which usually includes algorithms for processing images, optimization methods, and fault-tolerant mechanisms, etc.; the fault-tolerant mechanism refers to the error detection and correction mechanism possessed by the image stitching processing system, and when there are errors in the input error detection records, the system can automatically identify and repair these errors; the corrected data set refers to a data set processed by the fault-tolerant mechanism, which contains images or related information in which errors have been repaired; optionally, the correction of the error detection records can be achieved by utilizing the fault-tolerant mechanism in the image stitching processing system.

[0099] The present invention enables the system to process imperfect or noisy data in various splicing scenarios by extracting the corresponding correction factors in the correction data set, thereby improving the overall stability and applicability.

[0100] Among them, the correction factor refers to a group of factors used to correct and modify data, such as: offset correction factor, rotation correction factor and noise correction factor, etc. Optionally, the extraction of the corresponding correction factor in the corrected data set can be obtained through a machine learning algorithm, such as: linear regression, support vector machine and decision tree algorithms.

[0101] The present invention generates a corrected image set corresponding to the stitching candidate set based on the correction factor, which can improve image quality, geometric consistency and algorithm performance, and simplify the processing flow, providing a more reliable basis for subsequent image analysis and application.

[0102] The corrected image set refers to the corrected image set corresponding to the splicing candidate set obtained after image correction optimization processing.

[0103] As an embodiment of the present invention, generating a corrected image set corresponding to the stitching candidate set based on the correction factor includes: querying a factor data sequence corresponding to the correction factor; performing sequence parsing on the factor data sequence to obtain a plurality of correction parameters, the plurality of correction parameters including: a color shift parameter, a distortion correction parameter, and a brightness adjustment parameter; performing image correction optimization on the stitching candidate set based on the color shift parameter, the distortion correction parameter, and the brightness adjustment parameter to obtain a corrected image set corresponding to the stitching candidate set.

[0104] Among them, the factor data sequence refers to the data sequence corresponding to the correction factor; the multiple correction parameters refer to the correction parameters obtained by analyzing the factor data sequence, including color shift parameters, distortion correction parameters and brightness adjustment parameters; the color shift parameters refer to parameters used to adjust the color shift of the image, including hue, saturation, etc.; the distortion correction parameters refer to parameters used to correct distortion in the image, such as lens distortion, etc.; the brightness adjustment parameters refer to parameters used to adjust the brightness of the image to ensure that the overall exposure is appropriate.

[0105] Furthermore, the query of the factor data sequence corresponding to the correction factor can be obtained through a time series prediction model, such as an ARIMA or Prophet model; the sequence parsing of the factor data sequence can be obtained through a time series analysis method, such as the statsmodels library in Python; the image correction optimization of the splicing candidate set can be obtained through an image optimization tool, such as Photoshop, GIMP and other tools.

[0106] S5. Remove the stitching edges of the images in the corrected image set to obtain an edge-removed image set, perform color balance on the edge-removed image set to obtain a balanced image set, and use a preset particle swarm optimization algorithm to optimize and stitch the images in the balanced image set to obtain optimized stitched images corresponding to the images to be stitched.

[0107] The present invention obtains an edge-removed image set by removing the splicing edges of the images in the corrected image set, which can improve image quality, improve image analysis results, simplify subsequent processing procedures, enhance subsequent processing experience, and reduce data storage and transmission costs.

[0108] The edge-removed image set refers to a new image set obtained by removing the splicing edges of the images in the corrected image set using the extracted edge mask, that is, a set of images from which the edges have been removed.

[0109] As an embodiment of the present invention, the method of removing the stitched edges of the images in the corrected image set to obtain the edge-removed image set includes: performing gradient calculation on the images in the corrected image set to obtain a gradient image; performing threshold segmentation on the gradient image to obtain a binary image; performing morphological connection on the binary image to obtain a connected morphological image; extracting an edge mask from the connected morphological image; and using the edge mask to remove the stitched edges of the images in the corrected image set to obtain the edge-removed image set.

[0110] Among them, the gradient image refers to the image obtained after gradient calculation of the image in the corrected image set, which shows the intensity of changes in pixel values; the binary image refers to the image obtained after threshold segmentation is applied to the gradient image, in which only two pixel values ​​(usually black and white) are used to represent different parts of the image; the connection morphological image refers to the image obtained after morphological processing of the binary image, which is processed to connect edges and eliminate unnecessary noise points; the edge mask refers to the image extracted from the connection morphological image, which is used to mark the edge area and make a mask.

[0111] Furthermore, the gradient calculation of the images in the corrected image set can be achieved by the cv2.Sobel function in OpenCV; the threshold segmentation of the gradient image can be achieved by Otsu's binarization method; the morphological connection of the binary image can be achieved by the dilation (cv2.dilate) and erosion (cv2.erode) functions in OpenCV; and the removal of the spliced ​​edges of the images in the corrected image set can be achieved by utilizing the edge mask.

[0112] The present invention performs color balancing on the edge-removed image set to obtain a balanced image set, which is beneficial to improving visual effects, unifying styles, improving contrast, and reducing visual fatigue, thereby enhancing image quality and expressiveness.

[0113] The balanced image set refers to a set of images that have been processed with color balance.

[0114] As an embodiment of the present invention, color balancing the edge-removed image set to obtain a balanced image set includes: traversing the pixel points of the images in the edge-removed image set and calculating the average color values ​​corresponding to the pixel points; adjusting the contrast of the images in the edge-removed image set based on the average color value; matching the color saturation of the images in the edge-removed image set based on the contrast; and color balancing the edge-removed image set based on the color saturation to obtain a balanced image set.

[0115] Among them, the average color value refers to the average value of the color values ​​of the pixels in the image of the edge-removed image set; the contrast refers to the degree of difference between the colors or brightness in the images of the edge-removed image set; and the color saturation refers to the purity and vividness of the color.

[0116] Furthermore, the traversal of pixel points of the images in the edge-removed image set can be achieved by traversing and accessing the pixel points of the images through the OpenCV library in Python; the calculation of the average color value corresponding to the pixel point can be achieved through color space conversion, such as: Lab, HSV, etc.; the adjustment of the contrast of the images in the edge-removed image set can be achieved through histogram equalization, CLAHE and other technologies; the matching of the color saturation of the images in the edge-removed image set can be achieved through color enhancement or saturation adjustment algorithm; the color balancing of the edge-removed image set can be achieved through histogram equalization and color correction algorithm.

[0117] The present invention optimizes and splices the images in the balanced image set by using a preset particle swarm optimization algorithm to obtain an optimized spliced ​​image corresponding to the image to be spliced. This can improve the global search capability of the algorithm, accelerate the convergence speed, have strong adaptability, be easy to implement, and avoid falling into a local optimal solution, thereby obtaining a better image splicing effect.

[0118] Among them, the preset particle swarm optimization algorithm refers to an optimization algorithm based on swarm intelligence. The algorithm uses a group of randomly generated particles to represent the solution space of the problem to be optimized, and iteratively updates the position and velocity of each particle so that the entire group searches towards the global optimal solution; the optimized stitched image refers to the result image obtained by effectively stitching the images in the balanced image set through the particle swarm optimization algorithm. Optionally, the optimized stitching of the images in the balanced image set can be achieved by utilizing the preset particle swarm optimization algorithm.

[0119] First, the present invention obtains the images to be stitched, and uses a preset convolutional neural network to extract the feature descriptors corresponding to the images to be stitched, so as to learn the local and global features of the images, and the feature descriptors can more accurately represent the image content and enhance the features of the images, thereby improving the accuracy and stability of image stitching. At the same time, the present invention performs image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, thereby improving the quality and accuracy of image stitching, and can better preserve the information and features of the original images, while providing multi-view and consistent observation. The present invention generates a stitching candidate set corresponding to the images to be stitched based on the texture difference items, which can improve the stitching effect and improve automatic The present invention improves the degree of digitization, increases flexibility and reduces the stitching error rate, which has important benefits for image stitching tasks. Secondly, the present invention inputs the error detection record into a preset image stitching processing system, and then uses the fault-tolerant mechanism in the image stitching processing system to correct the error detection record to obtain a corrected data set, which can improve the accuracy, efficiency and robustness of the stitching result, while reducing the need for subsequent intervention, and can more robustly handle various stitching scenarios and improve overall stability. Furthermore, the present invention removes the stitching edges of the images in the corrected image set to obtain an edge-removed image set, which can improve image quality, improve image analysis results, simplify subsequent processing procedures, enhance subsequent processing experience, and reduce data storage and transmission costs. Therefore, the intelligent stitching method and system for electronic images based on a decision network proposed in the present invention can improve the stitching accuracy and efficiency of electronic images.

[0120] like Figure 2 , which is a functional module diagram of a system for intelligently splicing electronic images based on a decision network provided by an embodiment of the present invention.

[0121] The intelligent splicing system 200 for realizing electronic images based on a decision network of the present invention can be installed in an electronic device. According to the functions implemented, the intelligent splicing system 200 for realizing electronic images based on a decision network can include a feature similarity value calculation module 201, a texture difference item calculation module 202, an error detection module 203, an image correction module 204, and an optimization splicing module 205. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0122] In this embodiment, the functions of each module / unit are as follows:

[0123] The feature similarity value calculation module 201 is used to obtain the images to be spliced, extract the feature descriptors corresponding to the images to be spliced ​​using a preset convolutional neural network, and calculate the feature similarity values ​​corresponding to the images to be spliced ​​based on the feature descriptors;

[0124] The texture difference item calculation module 202 is used to perform image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, perform region segmentation on the registered image set using a preset clustering algorithm to obtain image segmentation regions, extract key feature points in the image segmentation regions, and query the saturation and color difference corresponding to the key feature points, and calculate the texture difference item corresponding to the image segmentation region based on the saturation and the color difference;

[0125] The error detection module 203 is used to generate a splicing candidate set corresponding to the to-be-spliced ​​images based on the texture difference item, and perform error detection on the splicing candidate set using a preset automatic detection mechanism to obtain an error detection record;

[0126] The image correction module 204 is used to input the error detection record into a preset image stitching processing system, correct the error detection record by using the fault tolerance mechanism in the image stitching processing system, obtain a corrected data set, extract the corresponding correction factor in the corrected data set, and generate a corrected image set corresponding to the stitching candidate set based on the correction factor;

[0127] The optimized stitching module 205 is used to remove the stitching edges of the images in the corrected image set to obtain an edge-removed image set, perform color balance on the edge-removed image set to obtain a balanced image set, and use a preset particle swarm optimization algorithm to optimize the stitching of the images in the balanced image set to obtain optimized stitched images corresponding to the images to be stitched.

[0128] In detail, the modules described in the intelligent splicing system 200 for realizing electronic images based on a decision network in the embodiment of the present invention adopt the same technical means as the intelligent splicing method for realizing electronic images based on a decision network described in the accompanying drawings when used, and can produce the same technical effects, which will not be repeated here.

[0129] like Figure 3 , which is a schematic diagram of the structure of an electronic device for realizing the intelligent splicing method of electronic images based on a decision network according to the present invention.

[0130] The electronic device may include a processor 30, a memory 31, a communication bus 32 and a communication interface 33, and may also include a computer program stored in the memory 31 and executable on the processor 30, such as an engineering safety supervision program based on artificial intelligence.

[0131] In some embodiments, the processor 30 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 30 is the control core (ControlUnit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 31 (for example, executing engineering safety supervision programs based on artificial intelligence, etc.), and calls data stored in the memory 31 to execute various functions of the electronic device and process data.

[0132] The memory 31 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 31 may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 31 may also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 31 may also include both an internal storage unit of the electronic device and an external storage device. The memory 31 may not only be used to store application software and various types of data installed in the electronic device, such as the code of the database configuration connection program, but also be used to temporarily store data that has been output or is to be output.

[0133] The communication bus 32 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 31 and at least one processor 30, etc.

[0134] The communication interface 33 is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0135] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0136] For example, although not shown, the electronic device may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor 30 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0137] It should be understood that the embodiment is for illustrative purposes only and the scope of the patent invention is not limited to this structure.

[0138] The database configuration connection program stored in the memory 31 in the electronic device is a combination of multiple computer programs. When running in the processor 30, it can achieve:

[0139] Acquire images to be stitched, extract feature descriptors corresponding to the images to be stitched using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be stitched based on the feature descriptors;

[0140] Based on the feature similarity values, the images to be stitched are registered to obtain a registered image set, the registered image set is segmented by a preset clustering algorithm to obtain an image segmentation region, key feature points in the image segmentation region are extracted, and the saturation and color difference corresponding to the key feature points are queried, and based on the saturation and the color difference, a texture difference item corresponding to the image segmentation region is calculated;

[0141] Based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images is generated, and an error detection is performed on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record;

[0142] After the error detection record is input into a preset image stitching processing system, the error detection record is corrected using the fault tolerance mechanism in the image stitching processing system to obtain a corrected data set, a corresponding correction factor in the corrected data set is extracted, and a corrected image set corresponding to the stitching candidate set is generated based on the correction factor;

[0143] The stitching edges of the images in the corrected image set are removed to obtain an edge-removed image set, the edge-removed image set is color balanced to obtain a balanced image set, and the images in the balanced image set are optimized and stitched using a preset particle swarm optimization algorithm to obtain optimized stitched images corresponding to the images to be stitched.

[0144] Specifically, the specific implementation method of the processor 30 for the above computer program can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0145] Furthermore, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium. The storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0146] The present invention further provides a storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0147] Acquire images to be stitched, extract feature descriptors corresponding to the images to be stitched using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be stitched based on the feature descriptors;

[0148] Based on the feature similarity values, the images to be stitched are registered to obtain a registered image set, the registered image set is segmented by a preset clustering algorithm to obtain an image segmentation region, key feature points in the image segmentation region are extracted, and the saturation and color difference corresponding to the key feature points are queried, and based on the saturation and the color difference, a texture difference item corresponding to the image segmentation region is calculated;

[0149] Based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images is generated, and an error detection is performed on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record;

[0150] After the error detection record is input into a preset image stitching processing system, the error detection record is corrected using the fault tolerance mechanism in the image stitching processing system to obtain a corrected data set, a corresponding correction factor in the corrected data set is extracted, and a corrected image set corresponding to the stitching candidate set is generated based on the correction factor;

[0151] The stitching edges of the images in the corrected image set are removed to obtain an edge-removed image set, the edge-removed image set is color balanced to obtain a balanced image set, and the images in the balanced image set are optimized and stitched using a preset particle swarm optimization algorithm to obtain optimized stitched images corresponding to the images to be stitched.

[0152] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0153] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0155] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0156] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0157] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0158] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in their embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. Intelligent splicing method of electronic images based on decision network, It is characterized in that The method comprises: Acquire images to be stitched, extract feature descriptors corresponding to the images to be stitched using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be stitched based on the feature descriptors; Based on the feature similarity values, the images to be stitched are registered to obtain a registered image set, the registered image set is segmented by a preset clustering algorithm to obtain an image segmentation region, key feature points in the image segmentation region are extracted, and the saturation and color difference corresponding to the key feature points are queried, and based on the saturation and the color difference, a texture difference item corresponding to the image segmentation region is calculated; Based on the texture difference item, a stitching candidate set corresponding to the to-be-stitched images is generated, and an error detection is performed on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record; After the error detection record is input into a preset image stitching processing system, the error detection record is corrected using the fault tolerance mechanism in the image stitching processing system to obtain a corrected data set, a corresponding correction factor in the corrected data set is extracted, and a corrected image set corresponding to the stitching candidate set is generated based on the correction factor; Removing the stitching edges of the images in the corrected image set to obtain an edge-removed image set, performing color balancing on the edge-removed image set to obtain a balanced image set, and optimizing and stitching the images in the balanced image set using a preset particle swarm optimization algorithm to obtain optimized stitching images corresponding to the images to be stitched; The calculating, based on the feature descriptors, feature similarity values ​​corresponding to the images to be stitched includes: Identify a descriptor dimension corresponding to the feature descriptor, and query an element scale corresponding to an element in the descriptor dimension; Based on the descriptor dimension and the element scale, the feature similarity value corresponding to the image to be stitched is calculated using the following formula: Wherein, Sd represents the feature similarity value corresponding to the image to be stitched, D refers to the descriptor dimension, i represents the element index in the descriptor dimension, and F 1 (i) represents the descriptor F 1 The value of the i-th element in F 2 (i) represents the descriptor F 2 The i-th element value in , S refers to the element scale; The descriptor dimension refers to the number of elements contained in the feature descriptor; the element scale refers to the ratio or scale of each element in the descriptor; The step of performing image registration on the images to be spliced ​​based on the feature similarity values ​​to obtain a registered image set includes: Based on the feature similarity value, query the spatial position corresponding to the feature element in the image to be stitched; Based on the spatial position, calculating a transformation matrix corresponding to the characteristic elements in the image to be stitched; Using the transformation matrix, the characteristic elements in the images to be spliced ​​are mapped to the same coordinates to obtain a characteristic element coordinate system; Based on the characteristic element coordinate system, the images to be stitched are registered to obtain a registered image set; The spatial position refers to the exact position coordinates of each feature element in the image to be stitched, which is represented by the pixel coordinates of the feature point; the transformation matrix refers to a matrix that describes the spatial relationship between feature elements; the feature element coordinate system refers to the transformation matrix obtained by applying calculations to map the feature elements in the image to be stitched into a common coordinate system so that they are aligned in space; The method of performing region segmentation on the registered image set by using a preset clustering algorithm to obtain image segmentation regions includes: Preprocessing the images in the registration image set to obtain preprocessed images; After extracting the image pixel values ​​in the preprocessed image, calculating the statistical features corresponding to the image pixel values; Randomly selecting k features from the statistical features as initial cluster centers, assigning each pixel in the preprocessed image to the nearest initial cluster center, and calculating the average pixel value in the initial cluster center; Based on the pixel average value, performing region segmentation on the registered image set to obtain image segmentation regions; The calculating, based on the saturation and the color difference, a texture difference item corresponding to the image segmentation area comprises: querying the regional feature points in the image segmentation area based on the saturation and the color difference; The texture difference item corresponding to the image segmentation area is calculated based on the regional feature points using the following formula: Wherein, T represents the texture difference item corresponding to the image segmentation area, j represents the index corresponding to the feature point of the area, n represents the total number of feature points of the area, S j It represents the saturation value of the feature point in the jth region, W j represents the weight of the jth regional feature point, D j Represents the color difference value of the jth region feature point; Generating a candidate splicing set corresponding to the to-be-spliced ​​image based on the texture difference item comprises: Calculating a similarity measure corresponding to the texture difference item; Based on the similarity measure, query the image sorting area corresponding to the texture difference item; Selecting the image with the highest similarity in the image sorting area as a candidate image; Combining the candidate images in the image selection area to obtain a candidate splicing set corresponding to the image to be spliced; The image sorting area refers to dividing the image to be stitched into multiple local areas; The step of generating a corrected image set corresponding to the stitching candidate set based on the correction factor comprises: Querying the factor data sequence corresponding to the correction factor; Sequence analysis is performed on the factor data sequence to obtain a plurality of correction parameters, wherein the plurality of correction parameters include: a color shift parameter, a distortion correction parameter, and a brightness adjustment parameter; Based on the color shift parameter, the distortion correction parameter, and the brightness adjustment parameter, performing image correction optimization on the stitching candidate set to obtain a corrected image set corresponding to the stitching candidate set; Removing the spliced ​​edges of the images in the corrected image set to obtain the edge-removed image set comprises: Performing gradient calculation on the images in the correction image set to obtain a gradient image; Performing threshold segmentation on the gradient image to obtain a binary image; Performing morphological connection on the binary images to obtain a connected morphological image; extracting an edge mask in the connected morphological image; Using the edge mask to remove the spliced ​​edges of the images in the corrected image set to obtain an edge-removed image set; The color balancing of the edge-removed image set to obtain a balanced image set comprises: Traversing the pixel points of the image in the edge-removed image set, and calculating the average color value corresponding to the pixel points; adjusting the contrast of the images in the edge-removed image set based on the average color value; matching the color saturation of the images in the edge-removed image set based on the contrast; Color balancing is performed on the edge-removed image set based on the color saturation to obtain a balanced image set.

2. Realize the intelligent splicing system of electronic images based on decision-making network, It is characterized in that The system is used to implement the intelligent stitching method of electronic images based on the decision network as claimed in claim 1, the system comprising: A feature similarity value calculation module is used to obtain the images to be spliced, extract feature descriptors corresponding to the images to be spliced ​​using a preset convolutional neural network, and calculate feature similarity values ​​corresponding to the images to be spliced ​​based on the feature descriptors; a texture difference item calculation module, configured to perform image registration on the images to be stitched based on the feature similarity values ​​to obtain a registered image set, perform region segmentation on the registered image set using a preset clustering algorithm to obtain image segmentation regions, extract key feature points in the image segmentation regions, query the saturation and color difference corresponding to the key feature points, and calculate the texture difference item corresponding to the image segmentation regions based on the saturation and the color difference; An error detection module, used to generate a stitching candidate set corresponding to the to-be-stitched images based on the texture difference item, and perform error detection on the stitching candidate set using a preset automatic detection mechanism to obtain an error detection record; An image correction module is used to input the error detection record into a preset image stitching processing system, and then use the fault tolerance mechanism in the image stitching processing system to correct the error detection record to obtain a corrected data set, extract the corresponding correction factor in the corrected data set, and generate a corrected image set corresponding to the stitching candidate set based on the correction factor; An optimization stitching module is used to remove the stitching edges of the images in the corrected image set to obtain an edge-removed image set, perform color balance on the edge-removed image set to obtain a balanced image set, and optimize and stitch the images in the balanced image set using a preset particle swarm optimization algorithm to obtain an optimized stitching image corresponding to the image to be stitched; The calculating, based on the feature descriptor, the feature similarity value corresponding to the image to be stitched comprises: Identify a descriptor dimension corresponding to the feature descriptor, and query an element scale corresponding to an element in the descriptor dimension; Based on the descriptor dimension and the element scale, the feature similarity value corresponding to the image to be stitched is calculated using the following formula: Wherein, Sd represents the feature similarity value corresponding to the image to be stitched, D refers to the descriptor dimension, i represents the element index in the descriptor dimension, and F 1 (i) represents the descriptor F 1 The value of the i-th element in F 2 (i) represents the descriptor F 2 The i-th element value in , S refers to the element scale; The descriptor dimension refers to the number of elements contained in the feature descriptor; the element scale refers to the ratio or scale of each element in the descriptor; The performing image registration on the images to be spliced ​​to obtain a registered image set based on the feature similarity value comprises: Based on the feature similarity value, query the spatial position corresponding to the feature element in the image to be stitched; Based on the spatial position, calculating a transformation matrix corresponding to the characteristic elements in the image to be stitched; Using the transformation matrix, the characteristic elements in the images to be spliced ​​are mapped to the same coordinates to obtain a characteristic element coordinate system; Performing image registration on the images to be stitched based on the characteristic element coordinate system to obtain a registered image set; The spatial position refers to the exact position coordinates of each feature element in the image to be stitched, which is represented by the pixel coordinates of the feature point; the transformation matrix refers to a matrix that describes the spatial relationship between feature elements; the feature element coordinate system refers to the transformation matrix obtained by applying calculations to map the feature elements in the image to be stitched into a common coordinate system so that they are aligned in space; The step of performing region segmentation on the registered image set using a preset clustering algorithm to obtain image segmentation regions comprises: Preprocessing the images in the registration image set to obtain preprocessed images; After extracting the image pixel values ​​in the preprocessed image, calculating the statistical features corresponding to the image pixel values; Randomly selecting k features from the statistical features as initial cluster centers, assigning each pixel in the preprocessed image to the nearest initial cluster center, and calculating the average pixel value in the initial cluster center; Performing region segmentation on the registered image set based on the pixel average value to obtain image segmentation regions; The calculating, based on the saturation and the color difference, a texture difference item corresponding to the image segmentation area comprises: querying the regional feature points in the image segmentation area based on the saturation and the color difference; Based on the regional feature points, the texture difference item corresponding to the image segmentation area is calculated using the following formula: Wherein, T represents the texture difference item corresponding to the image segmentation area, j represents the index corresponding to the feature point of the area, n represents the total number of feature points of the area, S j It represents the saturation value of the feature point in the jth region, W j represents the weight of the jth regional feature point, D j Represents the color difference value of the jth region feature point; Generating a candidate splicing set corresponding to the to-be-spliced ​​image based on the texture difference item comprises: Calculating a similarity measure corresponding to the texture difference item; Based on the similarity measure, query the image sorting area corresponding to the texture difference item; Selecting the image with the highest similarity in the image sorting area as a candidate image; Combining the candidate images in the image selection area to obtain a candidate splicing set corresponding to the image to be spliced; The image sorting area refers to dividing the image to be stitched into multiple local areas; Generating the corrected image set corresponding to the stitching candidate set based on the correction factor comprises: Querying the factor data sequence corresponding to the correction factor; Performing sequence analysis on the factor data sequence to obtain a plurality of correction parameters, the plurality of correction parameters including: a color shift parameter, a distortion correction parameter, and a brightness adjustment parameter; Based on the color shift parameter, the distortion correction parameter, and the brightness adjustment parameter, performing image correction optimization on the stitching candidate set to obtain a corrected image set corresponding to the stitching candidate set; The step of removing the spliced ​​edges of the images in the corrected image set to obtain the edge-removed image set includes: Performing gradient calculation on the images in the correction image set to obtain a gradient image; Performing threshold segmentation on the gradient image to obtain a binary image; Performing morphological connection on the binary images to obtain a connected morphological image; extracting an edge mask in the connected morphological image; Using the edge mask to remove the spliced ​​edges of the images in the corrected image set to obtain an edge-removed image set; The step of performing color balancing on the edge-removed image set to obtain a balanced image set includes: Traversing the pixel points of the image in the edge-removed image set, and calculating the average color value corresponding to the pixel points; Based on the average color value, adjusting the contrast of the images in the edge-removed image set; Based on the contrast, matching the color saturation of the images in the edge-removed image set; Based on the color saturation, color balancing is performed on the edge-removed image set to obtain a balanced image set.

Images