Image processing method, device, storage medium and electronic device

Abstract

The present specification discloses an image processing method, device, storage medium and electronic device. The method obtains a set of remote sensing image slices to be screened, which includes a plurality of remote sensing image slices to be screened, and determines the slice image features of each remote sensing image slice to be screened according to a preset feature extraction method. The set of remote sensing image slices to be screened is screened according to the slice image features to obtain an initial screening result. According to the initial screening result, a quality assessment result is obtained by using a pre-trained image quality assessment model. Based on the quality assessment result, a target remote sensing image slice is determined in the set of remote sensing image slices to be screened. The features of the remote sensing image slices are extracted, and a primary screening is performed according to the features. Then, according to the primary screening result, the remote sensing image slices are secondary screened using the image quality assessment model to obtain high-quality target remote sensing image slices. This not only improves the screening efficiency, but also improves the screening accuracy.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to an image processing method, device, storage medium and electronic device. Background Art

[0002] With the development of image processing technology, various types of imaging equipment can now be used to acquire remote sensing images and process them to obtain remote sensing image slices, which can then be used to perform business in areas such as environmental monitoring, resource surveys, urban planning, and agricultural analysis.

[0003] Due to the large amount of remote sensing image data and the influence of the environment during shooting, the quality of the remote sensing image slices may be low, such as the relatively simple features, low contrast, and artifacts in the remote sensing image slices. Low-quality slices not only fail to provide effective information for subsequent services, but also take up a lot of storage space, seriously affecting the execution effect of subsequent services. Therefore, the remote sensing image slice set can be manually screened to obtain high-quality remote sensing image slices before executing subsequent services. However, the efficiency of manual screening is extremely low, so how to improve the efficiency of screening remote sensing image slices is an urgent problem to be solved.

[0004] Based on this, this specification provides an image processing method. Summary of the invention

[0005] This specification provides an image processing method, device, storage medium and electronic device to partially solve the above problems existing in the prior art.

[0006] This manual adopts the following technical solutions:

[0007] This specification provides an image processing method, the method comprising:

[0008] Acquire a remote sensing image slice set to be screened, wherein the remote sensing image slice set to be screened includes a plurality of remote sensing image slices to be screened;

[0009] For each remote sensing image slice to be screened, determining the slice image features of the remote sensing image slice to be screened according to a preset feature extraction method;

[0010] According to the slice image features, the remote sensing image slice set to be screened is screened to obtain an initial screening result;

[0011] According to the initial screening results, a quality assessment result of the remote sensing image slice to be screened is obtained by using a pre-trained image quality assessment model;

[0012] Based on the quality assessment result, a target remote sensing image slice is determined in the remote sensing image slice set to be screened.

[0013] Optionally, the slice image feature includes at least one of a color ratio, a histogram feature, a contrast, and a gradient feature;

[0014] According to the preset feature extraction method, the slice image features of the remote sensing image slice to be screened are determined, specifically including:

[0015] Obtaining a preset color space of the remote sensing image slice to be screened, and determining a specified color ratio of the remote sensing image slice to be screened through the preset color space;

[0016] A grayscale image of the remote sensing image slice to be screened is obtained, and the contrast of the remote sensing image slice to be screened is determined according to the grayscale image; and a grayscale histogram of the remote sensing image slice to be screened is determined according to the grayscale image as a histogram feature of the remote sensing image slice to be screened.

[0017] Optionally, according to a preset feature extraction method, determining the slice image features of the remote sensing image slice to be screened specifically includes:

[0018] Determine the horizontal gradient and vertical gradient of the remote sensing image slice to be screened;

[0019] Determining the gradient amplitude of the remote sensing image slice to be screened according to the horizontal gradient and the vertical gradient;

[0020] According to the gradient amplitude, the gradient characteristics of the remote sensing image slice to be screened are determined.

[0021] Optionally, the remote sensing image slice set to be screened is screened according to the slice image features to obtain an initial screening result, specifically including:

[0022] For each remote sensing image slice to be screened, determining the similarity between the grayscale histogram of the remote sensing image slice to be screened and the grayscale histogram of a preset reference slice;

[0023] Determining whether to retain the remote sensing image slice to be screened according to at least one of the similarity, the specified color ratio, the contrast and the gradient feature;

[0024] The reserved remote sensing image slices to be screened are determined as initial screening results.

[0025] Optionally, according to the initial screening result, a quality assessment result of the remote sensing image slice to be screened is obtained by using a pre-trained image quality assessment model, specifically including:

[0026] According to the initial screening result, determining the remote sensing image slice to be evaluated in the remote sensing image slice set to be screened;

[0027] The remote sensing image slice to be evaluated is input into a pre-trained image quality assessment model to obtain a quality assessment score of the remote sensing image slice to be evaluated output by the image quality assessment model as a quality assessment result.

[0028] Optionally, inputting the remote sensing image slice to be evaluated into a pre-trained image quality assessment model to obtain a quality assessment score of the remote sensing image slice to be evaluated output by the image quality assessment model specifically includes:

[0029] Input the remote sensing image slice to be evaluated into a pre-trained image quality assessment model, and extract the features of the remote sensing image slice to be evaluated through deep convolution and point-by-point convolution;

[0030] Based on the features, a quality assessment score of the remote sensing image slice to be assessed is obtained and output.

[0031] Optionally, the method further comprises:

[0032] According to a preset optimization method, the target remote sensing image slice is optimized.

[0033] Optionally, training an image quality assessment model includes:

[0034] Obtain a sample remote sensing image slice and obtain a quality assessment result of the sample remote sensing image slice as a label;

[0035] Inputting the sample remote sensing image slice into an image quality assessment model to obtain a predicted assessment result output by the image quality assessment model;

[0036] The image quality assessment model is trained according to the predicted evaluation result, the label and a preset loss function, wherein the preset loss function includes at least one of a cross entropy loss function, an edge gradient loss function, a contrast loss function, a brightness unevenness loss function and a blur loss function.

[0037] This specification provides an image processing device, the device comprising:

[0038] A slice acquisition module is used to acquire a remote sensing image slice set to be screened, wherein the remote sensing image slice set to be screened includes a plurality of remote sensing image slices to be screened;

[0039] A feature determination module is used to determine the slice image features of each remote sensing image slice to be screened according to a preset feature extraction method;

[0040] An initial screening module, used to screen the remote sensing image slice set to be screened according to the slice image features to obtain an initial screening result;

[0041] A quality assessment module, used to obtain a quality assessment result of the remote sensing image slice to be screened based on the initial screening result and through a pre-trained image quality assessment model;

[0042] The target determination module is used to determine the target remote sensing image slice in the remote sensing image slice set to be screened based on the quality assessment result.

[0043] Optionally, the slice image feature includes at least one of a color ratio, a histogram feature, a contrast, and a gradient feature;

[0044] The feature determination module is specifically used to obtain a preset color space of the remote sensing image slice to be screened, and determine the specified color ratio of the remote sensing image slice to be screened through the preset color space; obtain a grayscale image of the remote sensing image slice to be screened, and determine the contrast of the remote sensing image slice to be screened based on the grayscale image; and determine the grayscale histogram of the remote sensing image slice to be screened based on the grayscale image as the histogram feature of the remote sensing image slice to be screened.

[0045] Optionally, the feature determination module is specifically used to determine the horizontal gradient and vertical gradient of the remote sensing image slice to be screened; determine the gradient amplitude of the remote sensing image slice to be screened based on the horizontal gradient and the vertical gradient; and determine the gradient feature of the remote sensing image slice to be screened based on the gradient amplitude.

[0046] Optionally, the initial screening module is specifically used to determine, for each remote sensing image slice to be screened, the similarity between the grayscale histogram of the remote sensing image slice to be screened and the grayscale histogram of a preset reference slice; determine whether to retain the remote sensing image slice to be screened based on the similarity, the specified color ratio, the contrast and at least one of the gradient features; and determine the retained remote sensing image slice to be screened as the initial screening result.

[0047] Optionally, the quality assessment module is specifically used to determine the remote sensing image slices to be evaluated in the set of remote sensing image slices to be screened based on the initial screening results; input the remote sensing image slices to be evaluated into a pre-trained image quality assessment model, and obtain the quality assessment score of the remote sensing image slices to be evaluated output by the image quality assessment model as the quality assessment result.

[0048] Optionally, the quality assessment module is specifically used to input the remote sensing image slice to be evaluated into a pre-trained image quality assessment model, extract the features of the remote sensing image slice to be evaluated through deep convolution and point-by-point convolution; based on the features, obtain the quality assessment score of the remote sensing image slice to be evaluated, and output it.

[0049] Optionally, the device further comprises:

[0050] The optimization module is used to optimize the target remote sensing image slice according to a preset optimization method.

[0051] Optionally, the device further comprises:

[0052] A training module is used to obtain a sample remote sensing image slice and obtain a quality assessment result of the sample remote sensing image slice as a label; input the sample remote sensing image slice into an image quality assessment model to obtain a predicted assessment result output by the image quality assessment model; and train the image quality assessment model according to the predicted assessment result, the label and a preset loss function, wherein the preset loss function includes at least one of a cross entropy loss function, an edge gradient loss function, a contrast loss function, a brightness unevenness loss function and a blur loss function.

[0053] This specification provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned image processing method is implemented.

[0054] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the image processing method is implemented when the processor executes the program.

[0055] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0056] It can be seen from the image processing method provided in this manual that this method extracts the features of remote sensing image slices, performs primary screening based on the features, and then uses the image quality assessment model to perform secondary screening on the remote sensing image slices based on the primary screening results to obtain high-quality target remote sensing image slices. This not only improves the screening efficiency, but also improves the screening accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0058] Figure 1 A flowchart of an image processing method provided in this specification;

[0059] Figure 2 A flow chart for determining slice image features provided in this specification;

[0060] Figure 3 Another image processing flow chart provided for this manual;

[0061] Figure 4 A schematic diagram of an image processing device provided in this specification;

[0062] Figure 5 A method corresponding to the Figure 1 Schematic diagram of the structure of an electronic device. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0064] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0065] Figure 1 A flowchart of an image processing method provided in this specification includes the following steps:

[0066] S100: Acquire a remote sensing image slice set to be screened, wherein the remote sensing image slice set to be screened includes a plurality of remote sensing image slices to be screened.

[0067] Generally speaking, the amount of data in remote sensing images is large, and the number of slices will also increase accordingly. If the remote sensing image includes a large area of ​​water, forest, etc., the slice features of the remote sensing image will be single, and the slices with single features have less information and are visually similar. For example, among 30 remote sensing image slices, 25 are slices of different areas of the same water area. Visually, these 25 slices have a high degree of similarity. Since the amount of information in the remote sensing image slices with single features is small, it will affect the effect of the subsequent business performed based on the remote sensing image slices, so it is necessary to screen the remote sensing image slices. In addition, when shooting remote sensing images, the remote sensing images may be blurred, artifacts, low contrast, etc. due to the influence of lighting, water vapor, dust and other particulate matter in the atmosphere affecting the propagation of light, resulting in low-quality remote sensing image slices. Low-quality remote sensing image slices will also affect the performance of business execution. Similarly, it is also necessary to screen the remote sensing image slices to obtain high-quality remote sensing image slices.

[0068] The remote sensing image set required to perform subsequent business can usually reach the PB level. The traditional manual screening method is extremely inefficient due to the huge amount of data. The existing automated processing technology mainly focuses on noise removal and simple image enhancement, and the accuracy of recognition and cleaning of low-quality remote sensing image slices is low. Therefore, this specification provides an image processing method. The execution subject of this specification can be an electronic device that can execute the solution of this specification, such as a server, a mobile phone, a personal computer (PC), a distributed cluster, etc., and can also be other computing devices with computing capabilities. This specification does not limit this. For the sake of convenience, this specification uses the server as the execution subject for description.

[0069] First, the server may first obtain remote sensing images through an image acquisition device, which may include remote sensing satellites, drones, etc., wherein the remote sensing image format may be GeoTIFF, JPEG, PNG, etc., which is not limited in this specification. Then, the image acquisition device sends the acquired remote sensing image to the server, and the server receives the remote sensing image sent by the image acquisition device.

[0070] Next, the server can divide the remote sensing image into several remote sensing image slices according to the requirements to obtain a set of remote sensing image slices to be screened. For example, the remote sensing image is cropped into several remote sensing image slices with a pixel size of 1024×1024. How to obtain remote sensing image slices from remote sensing images is a technology well known to those skilled in the art and will not be described in detail in this specification.

[0071] It can be understood that low-quality remote sensing image slices refer to remote sensing image slices with low contrast, single features, artifacts, and low clarity. In contrast, high-quality remote sensing image slices are determined relative to low-quality remote sensing image slices. It can also be understood that high-quality remote sensing image slices have high contrast, diverse features, no artifacts, and high clarity compared to low-quality remote sensing image slices.

[0072] S102: For each remote sensing image slice to be screened, determining the slice image features of the remote sensing image slice to be screened according to a preset feature extraction method.

[0073] Specifically, the slice image features include color features, histogram features, texture features, edge features, gradient features, etc., wherein the color features include color mean, variance, color ratio and hue distribution, etc., and the texture features include uniformity, contrast, energy, entropy, etc., and a specific feature combination can be flexibly selected according to different application scenarios. Other tangent image features may also be included, and this specification will not give examples one by one here, as long as the server can use the slice image features to execute step S104.

[0074] Color features can analyze each slice through RGB and HSV color space to calculate information such as color mean, variance, color ratio and hue distribution. Texture features use the gray level co-occurrence matrix (GLCM) algorithm to extract the texture features of the remote sensing image slices to be screened, including uniformity, contrast, energy, entropy, etc. Edge features use Canny edge detection technology to identify edge features in the remote sensing image slices to be screened, extract structural details in the remote sensing image slices to be screened, and gradient features are determined by Sobel operator, Scharr operator, etc.

[0075] Figure 2 The flowchart for determining the characteristics of the slice image provided in this specification is as follows: Figure 2 shown.

[0076] The server can first convert the remote sensing image slice to be screened into a grayscale image, that is, obtain the grayscale image of the remote sensing image slice to be screened. Then, based on the grayscale image, determine the contrast of the remote sensing image slice to be screened. Specifically, the server can first determine the pixel value of each pixel in the grayscale image and the grayscale mean of the grayscale image. Then, based on each pixel value and the grayscale mean, determine the grayscale standard deviation. Finally, based on the grayscale standard deviation, determine the contrast, and the grayscale standard deviation is positively correlated with the contrast. The grayscale standard deviation is as follows:

[0077]

[0078] in, is the pixel value of the grayscale image, is the grayscale mean.

[0079] After obtaining the grayscale image, the server can also determine the grayscale histogram of the remote sensing image slice to be screened based on the grayscale image as the histogram feature of the remote sensing image slice to be screened. Specifically, the server can convert the grayscale image into a grayscale histogram based on the histogram function. Its grayscale histogram As follows:

[0080]

[0081] in, is the histogram function of the grayscale image.

[0082] It should be noted that the server can determine the image features of some slices of the remote sensing image slices to be screened as needed for subsequent initial screening, and this specification does not impose any restrictions on this.

[0083] S104: Screening the remote sensing image slice set to be screened according to the slice image features to obtain an initial screening result.

[0084] In one or more embodiments of the present specification, during the initial screening, the slice image features of the remote sensing image slice to be screened may be compared with the slice image features of several preset low-quality slices to obtain the initial screening results.

[0085] The server can determine the similarity between the grayscale histogram of each remote sensing image slice to be screened and the grayscale histogram of the preset reference slice for each remote sensing image slice to be screened. The similarity can be expressed as follows:

[0086]

[0087] in, It is the grayscale histogram of the preset reference slice.

[0088] Whether to retain the remote sensing image slice to be screened is determined based on the similarity, the specified color ratio, the contrast and at least one of the gradient features, wherein the aforementioned similarity determination method is only an example of this description, and this description does not limit how to determine the similarity, and the preset reference slice is a preset low-quality slice.

[0089] When judging by similarity, when there is more than one preset reference slice. The server can determine the similarity between the grayscale histogram of the preset reference slice and the grayscale histogram of the remote sensing image slice to be screened for each preset reference slice. Among several similarities, the maximum similarity is determined. When the maximum similarity is less than the preset similarity threshold, it means that the remote sensing image slice to be screened is a high-quality slice, and the remote sensing image slice to be screened is retained. When there is only one preset reference slice, the server can only determine the similarity between the grayscale histogram of the preset reference slice and the grayscale histogram of the remote sensing image slice to be screened, and then determine whether to retain the remote sensing image slice to be screened.

[0090] Similarly, the server can also obtain and determine the specified color ratio, contrast and gradient characteristics of the preset reference slice, and then the server can determine whether to retain the remote sensing image slice to be screened based on at least one of the similarity, the specified color ratio, the contrast and the gradient characteristics, and determine the retained remote sensing image slice to be screened as the initial screening result. For example, it is determined whether the specified color ratio of the remote sensing image slice to be screened is greater than the specified color ratio of the preset reference slice. If so, the remote sensing image slice to be screened is not retained. The contrast and gradient characteristics are similar to the aforementioned similarity and specified color ratio, and will not be repeated here.

[0091] Of course, the quality classification result of the remote sensing image slice to be screened can also be determined based on at least one of the similarity, the specified color ratio, the contrast and the gradient feature, and marked as the initial screening result. Among them, the quality classification results include low quality and high quality. Specifically, preset thresholds can be set for each slice image feature, which are respectively a similarity threshold, a specified color ratio threshold, a contrast threshold and a gradient feature threshold.

[0092] For example, if the contrast is less than the preset contrast threshold, the quality type of the remote sensing image slice to be screened is low quality, and if the specified color ratio is not greater than the preset specified color ratio threshold, the quality type of the remote sensing image slice to be screened is high quality. Whether to delete the remote sensing image slice to be screened of low quality can be executed according to the needs of subsequent business, and this manual does not limit this.

[0093] S106: According to the initial screening result, a quality assessment result of the remote sensing image slice to be screened is obtained by using a pre-trained image quality assessment model.

[0094] Since there are misjudgments in the initial screening results, for example, low-quality slices are determined to be high-quality slices, the server can further screen through a pre-trained image quality assessment model to improve the accuracy of the screening.

[0095] Specifically, the server determines the remote sensing image slices to be evaluated in the remote sensing image slice set to be screened based on the initial screening results. Generally speaking, in the second screening, only the high-quality slices in the initial screening results are screened for the second time. Therefore, based on the initial screening results, the server determines the remote sensing image slices to be screened whose initial screening results are high-quality slices in the remote sensing image slice set to be screened as the remote sensing image slices to be evaluated. Of course, in order to avoid the situation where high-quality slices are determined as low-quality slices in the initial screening, all remote sensing image slices to be screened in the remote sensing image slice set to be screened can also be determined as remote sensing image slices to be evaluated.

[0096] The server can input the remote sensing image slice to be evaluated into the pre-trained image quality assessment model, and extract the features of the remote sensing image slice to be evaluated through deep convolution and point-by-point convolution. Each feature reflects different aspects of the image and provides basic data for subsequent quality assessment. The convolution formula is as follows:

[0097]

[0098]

[0099] Among them, Depth wise Convolution is depth convolution, Point wise Convolution is point-wise convolution, X is the input remote sensing image slice to be evaluated, W is the depth convolution kernel, P is the point-wise convolution kernel, Y is the intermediate feature map, and Z is the final output feature map. i and j are pixel coordinates, k and l are the scales (steps) of the convolution kernel sliding, and m and n are the convolution positions of the convolution kernel.

[0100] Finally, based on the feature, the quality assessment score of the remote sensing image slice to be evaluated is obtained and output.

[0101] S108: Based on the quality assessment result, determine a target remote sensing image slice in the remote sensing image slice set to be screened.

[0102] The server can classify the quality of the remote sensing image slice to be evaluated according to the quality evaluation score. Specifically, it can determine whether the remote sensing image slice to be screened corresponding to the quality evaluation score is a high-quality remote sensing image slice according to the quality evaluation score and a preset score threshold. If so, the remote sensing image slice to be screened is determined as the target remote sensing image slice. Afterwards, the server can save the target remote sensing image slice to a specified data space, and save the remote sensing image slice set to be screened except the target remote sensing image slice to another specified data space.

[0103] Since in step S106, the server can determine all the remote sensing image slices to be screened in the remote sensing image slice set to be screened as the remote sensing image slices to be evaluated, there may be differences between the initial screening results of the remote sensing image slices to be screened and the quality evaluation results of the model output. Then, the remote sensing image slices to be screened can be displayed to the user, and the user makes a decision to determine the remote sensing image slices to be screened. That is to say, based on the quality evaluation results, in the remote sensing image slice set to be screened, the slices to be displayed are determined, and in response to the user's quality classification operation, the quality classification results of the slices to be displayed are determined, and the slices to be displayed with a quality classification result of high quality are determined as the target remote sensing image slices.

[0104] It should be noted that user decision is limited to the case where the initial screening result of the remote sensing image slice to be screened is a low-quality slice, and the quality assessment result output by the model is a high-quality slice. If the initial screening result of the remote sensing image slice to be screened is a high-quality slice, and the quality assessment result output by the model is a low-quality slice, no user decision is required.

[0105] based on Figure 1The image processing method shown in the figure obtains a set of remote sensing image slices to be screened, which includes several remote sensing image slices to be screened, and determines the slice image features of each remote sensing image slice to be screened according to a preset feature extraction method. According to the slice image features, the remote sensing image slice set to be screened is screened to obtain an initial screening result. According to the initial screening result, a quality assessment result is obtained by using a pre-trained image quality assessment model. Based on the quality assessment result, the target remote sensing image slice is determined in the remote sensing image slice set to be screened. By extracting the features of the remote sensing image slices, a primary screening is performed according to the features, and then according to the primary screening results, the remote sensing image slices are secondary screened using the image quality assessment model to obtain high-quality target remote sensing image slices. This not only improves the screening efficiency, but also improves the screening accuracy.

[0106] With respect to step S100, after the server obtains the remote sensing image, it may also preprocess the remote sensing image. The preprocessing includes denoising, geometric correction and cropping, etc. The denoising method may include median filtering, bilateral filtering, etc. Preprocessing the remote sensing image can make the size of the remote sensing image uniform and ensure data consistency.

[0107] For step S102, when determining the color ratio of the remote sensing image slice to be screened, the server can obtain the preset color space of the remote sensing image slice to be screened, including the HSV color space, where HSV is hue, saturation, and value. Then, through the preset color space, the specified color ratio of the remote sensing image slice to be screened is determined, and the specified color can be set as needed. For example, in agricultural analysis, the specified color can be green, and the green ratio of the remote sensing image slice to be screened can be determined to determine the proportion of plant species. For another example, the forest coverage can be determined by the green ratio, and the green ratio is positively correlated with the proportion of plant species and forest coverage.

[0108] Specifically, the server can obtain the preset color space based on the RGB (Red, Green, Blue) image of the remote sensing image slice to be screened, and then use the RGB image and the known conversion relationship to determine the specified color ratio. It can also use existing tools to directly determine the HSV color space. This description does not limit how to determine the preset color space. After obtaining the HSV color space, the ratio of the specified color pixels can be determined by specifying a color pixel mask to obtain the specified color ratio. Taking green as an example, the ratio of green color pixels is as follows:

[0109]

[0110] in, is the green pixel mask, (i, j) is the position of the mask, is the resolution of the remote sensing image slice to be screened. It can also be expressed as ColorMask, which means that any color can be specified.

[0111] When determining the gradient feature, the horizontal gradient and vertical gradient of the remote sensing image slice to be screened can be determined first. For example, the horizontal and vertical gradients of the remote sensing image slice to be screened can be determined by using the Sobel operator, as shown in the following formula:

[0112]

[0113] in, is the horizontal gradient, Of course, the horizontal gradient and vertical gradient of the remote sensing image slice to be screened can also be determined by other methods, and this specification does not limit this.

[0114] Then, the gradient amplitude of the remote sensing image slice to be screened can be determined according to the horizontal gradient and the vertical gradient. As shown in the following formula:

[0115]

[0116] in, is the gradient amplitude.

[0117] In addition, the server also needs to determine the gradient mean of the remote sensing image slice to be screened. Finally, the server can determine the gradient characteristics of the remote sensing image slice to be screened based on the gradient amplitude and the gradient mean. As shown in the following formula:

[0118]

[0119] in, is the mean gradient, It is a gradient feature and can be used to evaluate the detail richness of the remote sensing image slices to be screened.

[0120] After determining the slice image features, the server can also perform standardization and dimensionality reduction on the determined slice image features, such as using Z-score standardization or Min-Max normalization methods to eliminate the dimensional differences between different features. The server can use principal component analysis PCA and recursive feature elimination (RFE) to screen the standardized features, retain features that are highly related to slice quality, and reduce redundant data through dimensionality reduction technology to improve computing efficiency, thereby improving the efficiency and accuracy of subsequent screening.

[0121] For steps S102 to S104, in order to further save computing resources and improve screening efficiency, the calculation can be performed according to the difficulty of calculating the slice image features, and then the screening can be performed. Generally speaking, the calculation difficulty of color ratio, contrast, histogram features, and gradient features increases in sequence. Then, the server can first determine the color ratio of each remote sensing image slice to be screened, and perform a first screening based on the color ratio to obtain the first screening result. Then determine the contrast of the high-quality slices in the first screening result, and perform a second screening based on the contrast to obtain the second screening result. Then determine the histogram features of the high-quality slices in the second screening result, and perform a third screening based on the histogram features to obtain the third screening result. Finally, determine the gradient features of the high-quality slices in the third screening result, and perform a fourth screening based on the gradient features to obtain the initial screening result.

[0122] In addition, under the premise that computing resources and computing time permit, the similarity between the grayscale histograms of any two remote sensing image slices to be screened in the remote sensing image slice set to be screened can be determined. When the similarity is greater than the preset similarity threshold, only one of the remote sensing image slices to be screened is retained.

[0123] For step S106, when the image quality assessment model is used for secondary screening, in the convolution process, not only can features be extracted, but also the features can be fused in combination with a multi-layer perceptron to obtain fused features. Specifically, the image quality assessment model extracts multi-scale features of the remote sensing image slice to be evaluated through deep convolution, point-by-point convolution, and ReLU activation function, and generates fused features based on a multi-layer perceptron. Through the global average pooling layer, the average value of the fused features obtained by each convolution is used as the quality assessment score. That is, based on the fused feature, the quality assessment score of the remote sensing image slice to be evaluated is obtained and output to obtain the quality assessment result.

[0124] In addition, in order to prevent overfitting and improve the performance of the model on unknown data, regularization methods such as Dropout and Batch Normalization are added to the model to further improve the accuracy of the quality assessment results output by the image quality assessment model.

[0125] The image quality assessment model in this description can use a lightweight MobileNet convolutional neural network model, then the quality assessment score is as follows:

[0126]

[0127] This lightweight model is designed based on the architecture of deep separable convolution. By decomposing the traditional convolution into deep convolution and point-by-point convolution, the number of parameters and computation is significantly reduced, further improving the screening efficiency. In other words, by adopting deep separable convolution, MobileNet maintains the ability to capture complex image features while reducing computational costs. When processing large-scale remote sensing image slices, the model can significantly increase classification speed and reduce memory usage. Compared with traditional convolutional neural networks, such as VGG and ResNet, MobileNet's number of parameters and computation are reduced by about 70% and 90%, respectively, thereby achieving more efficient slice screening.

[0128] With respect to step S108, after the server obtains the target remote sensing image slice, it can also optimize the target remote sensing image slice according to a preset optimization method and then store it. The optimization method may include image enhancement, denoising, etc. to improve the overall quality of the image. Laplace enhancement, gamma correction and other technologies may also be used to further optimize the target remote sensing image slice to enhance the visual effect and information content of the target remote sensing image slice for subsequent business execution. Finally, the optimized target remote sensing image slice is stored.

[0129] During storage, the server can store the selected and optimized remote sensing image slices back into the database to replace the original slices. The database can use distributed storage technology to support efficient access and management of large-scale image data.

[0130] In addition, the slices stored in the database can be updated regularly, and new high-quality image slices can be added and outdated or useless slices can be removed according to actual application needs, ensuring that the image data in the database always maintains high quality and high practicality.

[0131] Of course, low-quality image slices can also be stored and used for subsequent image quality assessment model optimization or re-screening.

[0132] In order to improve the execution efficiency, steps S100 to S108 of this specification can be executed in multiple threads. Specifically, the server can process the remote sensing image slices to be screened in parallel through multi-threading technology to improve the processing efficiency, and the maximum number of threads is automatically configured according to the number of system CPU cores. The remote sensing image slices to be screened can also be processed in batches, and the number of image slices in each batch can be dynamically adjusted according to the size of the data set to achieve a balance between processing efficiency and memory usage.

[0133] It should be noted that the remote sensing image slices to be screened can also be directly input into the image quality assessment model to obtain the quality assessment results, and then obtain the target remote sensing image slices. Only executing the aforementioned steps can also improve the screening efficiency.

[0134] This specification also provides a training method for an image quality assessment model. Specifically, the server can obtain a sample remote sensing image slice and obtain the quality assessment result of the sample remote sensing image slice as a label. The sample remote sensing image slice can be a sample image slice of various qualities to ensure that the model can identify different types of low-quality slices, such as low contrast, single features, severe artifacts, etc. In addition, the sample remote sensing image slices can come from a variety of different scenes to enhance the generalization ability of the model and make the image quality assessment model suitable for a variety of application environments.

[0135] The sample remote sensing image slice is input into the image quality assessment model to obtain the predicted assessment result output by the image quality assessment model. The image quality assessment model is trained according to the predicted assessment result, the label and the preset loss function. The preset loss function includes at least one of a cross entropy loss function, an edge gradient loss function, a contrast loss function, a brightness unevenness loss function and a blur loss function.

[0136] This is because high-quality slices usually have clear edges, better contrast, and clearer boundaries between objects. In addition, images with uneven brightness are often associated with low quality, and blurred images are generally considered to be low-quality images. Therefore, in order to further improve the accuracy of the image quality assessment model, the cross entropy loss function, edge gradient loss function, contrast loss function, uneven brightness loss function and blur loss function are added to this specification. Based on at least one of the edge gradient loss function, contrast loss function, uneven brightness loss function and blur loss function, an auxiliary loss can be determined. The loss function in this specification is as follows:

[0137]

[0138]

[0139] in, is the cross entropy loss for quality classification, is an auxiliary loss used to enhance the model’s sensitivity to high-quality slice features. Edge gradient loss is used to preserve the edge details of the slice image and ensure clear edges. is the contrast loss, used to ensure that the slice has appropriate contrast. Brightness non-uniformity loss is used to reduce the phenomenon of inconsistent brightness in the slice. is the blur loss, used to reduce blur in the slice.

[0140] also, , , and It is a hyperparameter used to balance the weights of different loss items. It can be set as needed, such as setting them all to 0.1. This manual does not limit this.

[0141] During the training process, the server can optimize the model parameters of the image quality assessment model through the back-propagation algorithm, and adjust hyperparameters such as learning rate and batch size in combination with cross-validation technology to ensure the stability and accuracy of the image quality assessment model.

[0142] After the image quality assessment model is trained, the server can also regularly evaluate the screening effect of the model, using a variety of indicators such as accuracy, recall rate, F1 score, etc. to detect the performance of the image quality assessment model in remote sensing image slices in different scenarios to ensure its reliability in practical applications. According to the performance evaluation results, the image quality assessment model is continuously optimized and updated. For example, new training data is introduced, the model structure of the image quality assessment model is adjusted, and new features are introduced to continuously improve the screening accuracy and processing efficiency of the image quality assessment model to adapt to the dynamic changes and diversified needs of remote sensing image data.

[0143] Example:

[0144] Figure 3 Another image processing flow chart provided for this manual is as follows: Figure 3 shown.

[0145] The server can first obtain the original remote sensing images from remote sensing satellites or drone images, and all the acquired remote sensing images will be stored in the original slice storage database. The database is organized by province, city or specific area, and stores the relevant remote sensing images and their corresponding slice data for each province. Figure 3 Includes remote sensing images of Province 1 and Province 2.

[0146] After that, the original remote sensing image is preprocessed, cropped into remote sensing image slices of preset size and stored separately. Each slice will be accompanied by metadata, including image source, slice number, location coordinates, generation time and other information to ensure the convenience of subsequent screening and management. Figure 3 In the , each remote sensing image has a corresponding slice.

[0147] Then, for each remote sensing image slice, we screen it based on multi-feature analysis, including color, texture, edge, etc., to evaluate the quality of the remote sensing image slice. We can also evaluate the quality of the remote sensing image slice through techniques such as histogram distribution comparison, green percentage calculation, and gradient standard deviation. Low-quality remote sensing image slices will be marked as low-quality slices.

[0148] Next, a model trained with lightweight deep neural network models such as MobileNet is used to further evaluate the quality of the slices to be tested, which have been screened based on multi-feature analysis. The lightweight deep neural network model outputs a quality score based on the high-dimensional features of the remote sensing image slices. Remote sensing image slices with scores above a preset threshold can be marked as "high-quality slices", and remote sensing image slices with scores below the threshold can be marked as low-quality slices.

[0149] Furthermore, the selected high-quality slices are stored in the high-quality slice set and organized into the slice quality screening database of each province. High-quality slices can be used for subsequent remote sensing analysis, urban planning, agricultural monitoring and other applications. The low-quality slices screened out are archived in the low-quality slice set for future optimization training of lightweight deep neural network models. Low-quality slices will also be specially marked in the database so that users can easily identify and choose whether to perform subsequent processing or eliminate them.

[0150] In order to cope with the processing of large-scale remote sensing data, a multi-threaded parallel strategy is used for screening. Image slices are processed in parallel by region or image source, which greatly improves processing efficiency and shortens screening time.

[0151] In addition, to ensure the scalability and efficiency of the system, a distributed database is used to manage and store slice data. The databases in each area are independent of each other, while sharing the core screening algorithm to ensure that slices in different areas can update the screening results synchronously.

[0152] The server can also visualize the screening results of high-quality and low-quality slices through the user interface, allowing users to view the quality score, metadata and screening details of each slice. Users can customize the screening conditions, such as quality thresholds and feature weights, through the interface and view changes in the screening results in real time.

[0153] The server can also automatically update the slice data in the database based on newly added remote sensing images and regularly re-screen existing slices to ensure that the high-quality slice set in the database is always kept up to date.

[0154] In summary, the image processing method provided in this specification can quickly and accurately identify low-quality slices from millions of slices, such as low contrast, blur, hue distortion, and repeated textures, to ensure the diversity of remote sensing image slices, while further improving the overall quality of remote sensing image datasets. It can also be understood that the image processing method provided in this specification can significantly reduce the proportion of slice data with low image quality and low information content in remote sensing datasets and ensure the diversity of slices, improve the practical application value of remote sensing image slices, and can be used for remote sensing image data processing and the construction of large-scale high-quality remote sensing datasets.

[0155] The above is one or more implementation methods of this specification, based on Figure 1 The present specification also provides a corresponding image processing device, such as Figure 4 shown.

[0156] Figure 4 A schematic diagram of an image processing device provided in this specification includes:

[0157] The slice acquisition module 400 is used to acquire a remote sensing image slice set to be screened, wherein the remote sensing image slice set to be screened includes a plurality of remote sensing image slices to be screened;

[0158] The feature determination module 402 is used to determine the slice image features of each remote sensing image slice to be screened according to a preset feature extraction method;

[0159] An initial screening module 404 is used to screen the remote sensing image slice set to be screened according to the slice image features to obtain an initial screening result;

[0160] The quality assessment module 406 is used to obtain the quality assessment result of the remote sensing image slice to be screened according to the initial screening result through a pre-trained image quality assessment model;

[0161] The target determination module 408 is used to determine a target remote sensing image slice in the remote sensing image slice set to be screened based on the quality assessment result.

[0162] Optionally, the slice image feature includes at least one of a color ratio, a histogram feature, a contrast, and a gradient feature;

[0163] The feature determination module 402 is specifically used to obtain a preset color space of the remote sensing image slice to be screened, and determine the specified color ratio of the remote sensing image slice to be screened through the preset color space; obtain a grayscale image of the remote sensing image slice to be screened, and determine the contrast of the remote sensing image slice to be screened based on the grayscale image; and determine the grayscale histogram of the remote sensing image slice to be screened based on the grayscale image as the histogram feature of the remote sensing image slice to be screened.

[0164] Optionally, the feature determination module 402 is specifically used to determine the horizontal gradient and vertical gradient of the remote sensing image slice to be screened; determine the gradient amplitude of the remote sensing image slice to be screened based on the horizontal gradient and the vertical gradient; and determine the gradient feature of the remote sensing image slice to be screened based on the gradient amplitude.

[0165] Optionally, the initial screening module 404 is specifically used to determine, for each remote sensing image slice to be screened, the similarity between the grayscale histogram of the remote sensing image slice to be screened and the grayscale histogram of a preset reference slice; determine whether to retain the remote sensing image slice to be screened based on the similarity, the specified color ratio, the contrast and at least one of the gradient features; and determine the retained remote sensing image slice to be screened as the initial screening result.

[0166] Optionally, the quality assessment module 406 is specifically used to determine the remote sensing image slices to be evaluated in the remote sensing image slice set to be screened according to the initial screening results; input the remote sensing image slices to be evaluated into a pre-trained image quality assessment model, and obtain the quality assessment score of the remote sensing image slices to be evaluated output by the image quality assessment model as the quality assessment result.

[0167] Optionally, the quality assessment module 406 is specifically used to input the remote sensing image slice to be evaluated into a pre-trained image quality assessment model, extract the features of the remote sensing image slice to be evaluated through deep convolution and point-by-point convolution; based on the features, obtain the quality assessment score of the remote sensing image slice to be evaluated, and output it.

[0168] Optionally, the device further comprises:

[0169] The optimization module is used to optimize the target remote sensing image slice according to a preset optimization method.

[0170] Optionally, the device further comprises:

[0171] A training module is used to obtain a sample remote sensing image slice and obtain a quality assessment result of the sample remote sensing image slice as a label; input the sample remote sensing image slice into an image quality assessment model to obtain a predicted assessment result output by the image quality assessment model; and train the image quality assessment model according to the predicted assessment result, the label and a preset loss function, wherein the preset loss function includes at least one of a cross entropy loss function, an edge gradient loss function, a contrast loss function, a brightness unevenness loss function and a blur loss function.

[0172] This specification also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 An image processing method is provided.

[0173] This manual also provides Figure 5 The one shown corresponds to Figure 1 Schematic diagram of the structure of an electronic device. Figure 5As shown, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 5 The image processing method.

[0174] Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the executor of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0175] In the 1990s, it was very clear whether the improvement of a technology was hardware improvement (for example, improvement of the circuit structure of diodes, transistors, switches, etc.) or software improvement (improvement of the method flow). However, with the development of technology, many improvements of the method flow today can be regarded as direct improvements of the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be implemented with hardware entity modules. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask chip manufacturers to design and make dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0176] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0177] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0178] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0179] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0180] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0181] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0182] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0183] In a typical configuration, a computing device includes one or more processors (first chips), input / output interfaces, network interfaces, and memory.

[0184] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0185] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0186] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0187] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0189] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0190] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. An image processing method, characterized in that: The method comprises: Acquire a remote sensing image slice set to be screened, wherein the remote sensing image slice set to be screened includes a plurality of remote sensing image slices to be screened; For each remote sensing image slice to be screened, determining the slice image features of the remote sensing image slice to be screened according to a preset feature extraction method; According to the slice image features, the remote sensing image slice set to be screened is screened to obtain an initial screening result; According to the initial screening results, a quality assessment result of the remote sensing image slice to be screened is obtained by using a pre-trained image quality assessment model; Based on the quality assessment result, determining a target remote sensing image slice in the remote sensing image slice set to be screened; Wherein, the slice image feature includes at least one of color ratio, histogram feature, contrast, and gradient feature; According to the slice image features, the remote sensing image slice set to be screened is screened to obtain an initial screening result, which specifically includes: For each remote sensing image slice to be screened, determining the similarity between the grayscale histogram of the remote sensing image slice to be screened and the grayscale histogram of a preset reference slice; Determining whether to retain the remote sensing image slice to be screened according to at least one of the similarity, the color ratio, the contrast and the gradient feature; The reserved remote sensing image slices to be screened are determined as initial screening results; According to the initial screening results, the quality assessment results of the remote sensing image slices to be screened are obtained through a pre-trained image quality assessment model, specifically including: According to the initial screening result, determining the remote sensing image slice to be evaluated in the remote sensing image slice set to be screened; Inputting the remote sensing image slice to be evaluated into a pre-trained image quality assessment model, and obtaining a quality assessment score of the remote sensing image slice to be evaluated output by the image quality assessment model as a quality assessment result; Inputting the remote sensing image slice to be evaluated into a pre-trained image quality assessment model to obtain a quality assessment score of the remote sensing image slice to be evaluated output by the image quality assessment model specifically includes: Input the remote sensing image slice to be evaluated into a pre-trained image quality assessment model, and extract the features of the remote sensing image slice to be evaluated through deep convolution and point-by-point convolution; Based on the features, a quality assessment score of the remote sensing image slice to be assessed is obtained and output.

2. The method according to claim 1, characterized in that According to the preset feature extraction method, the slice image features of the remote sensing image slice to be screened are determined, specifically including: Obtaining a preset color space of the remote sensing image slice to be screened, and determining a specified color ratio of the remote sensing image slice to be screened through the preset color space; A grayscale image of the remote sensing image slice to be screened is obtained, and the contrast of the remote sensing image slice to be screened is determined according to the grayscale image; and a grayscale histogram of the remote sensing image slice to be screened is determined according to the grayscale image as a histogram feature of the remote sensing image slice to be screened.

3. The method according to claim 1, characterized in that According to the preset feature extraction method, the slice image features of the remote sensing image slice to be screened are determined, specifically including: Determine the horizontal gradient and vertical gradient of the remote sensing image slice to be screened; Determining the gradient amplitude of the remote sensing image slice to be screened according to the horizontal gradient and the vertical gradient; According to the gradient amplitude, the gradient characteristics of the remote sensing image slice to be screened are determined.

4. The method according to claim 1, characterized in that The method further comprises: According to a preset optimization method, the target remote sensing image slice is optimized.

5. The method according to claim 1, characterized in that Train the image quality assessment model, including: Obtain a sample remote sensing image slice and obtain a quality assessment result of the sample remote sensing image slice as a label; Inputting the sample remote sensing image slice into an image quality assessment model to obtain a predicted assessment result output by the image quality assessment model; The image quality assessment model is trained according to the predicted evaluation result, the label and a preset loss function, wherein the preset loss function includes at least one of a cross entropy loss function, an edge gradient loss function, a contrast loss function, a brightness unevenness loss function and a blur loss function.

6. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1 to 5 is implemented.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method described in any one of claims 1 to 5 is implemented.

Images