An optical image cloud and fog complexity recognition method

Abstract

The application discloses a kind of optical image cloud fog complexity identification methods, belong to image cloud fog complexity identification technical field;Including the following steps S1, to the optical image to be detected is confirmed to extreme cloud fog scene;S2, after completing S1, determine the cloud fog pixel area of optical image to be detected, complete the determination of certainty cloud fog area, certainty non-cloud fog area and non-determination area determination;S3, connect the certainty cloud fog area in the optical image to be detected, obtain cloud fog area structure feature, form preliminary pixel-level cloud fog mask;S4, further determination is carried out to non-determination area, and it is spliced with cloud fog area structure feature, complete optical image cloud fog complexity identification;The application adopts the above method, and the progressive identification structure of extreme scene judgment, morphological structure feature extraction judgment and threshold segmentation supplementary judgment is designed, the accurate identification of optical image cloud fog area pixel level is realized, and the accurate perception ability of optical image is improved.

Description

Technical Field

[0001] This invention belongs to the field of image cloud and fog complexity identification technology, and in particular relates to a method for identifying cloud and fog complexity in optical images. Background Technology

[0002] Currently, in applications such as optical image perception and remote sensing image perception, there are problems such as inconsistent image quality and difficulty in accurately removing cloud and fog regions, which greatly limits the ability to accurately identify targets. Existing cloud and fog state identification methods include threshold segmentation, texture analysis, and deep learning. Among them, threshold segmentation is difficult to adapt directly to complex scenes, texture analysis has insufficient ability to identify thin and fragmented clouds, and deep learning methods require pixel-level labeled data, which consumes a lot of human resources. In existing technologies, morphological methods are often used for the finishing and edge optimization of cloud and fog regions, usually as a post-processing step after threshold segmentation, failing to fully utilize the advantages of morphology in extracting structural features of cloud and fog regions. Summary of the Invention

[0003] The purpose of this invention is to provide a method for identifying the complexity of cloud and fog in optical images. It designs a progressive identification structure that includes extreme scene judgment, morphological structural feature extraction judgment, and threshold segmentation supplementary judgment, which realizes pixel-level accurate identification of cloud and fog regions in optical images and supports the improvement of the accurate perception capability of optical images.

[0004] To achieve the above objectives, the present invention provides a method for identifying the complexity of cloud and fog in optical images, comprising the following steps: S1. Confirm the extreme cloud and fog scene of the optical image to be tested; S2. After completing S1, determine the cloud and fog pixel region of the optical image to be detected, and complete the determination of deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region. S3. Connect the deterministic cloud and fog regions in the optical image to be detected to obtain the structural features of the cloud and fog regions and form a preliminary pixel-level cloud and fog mask. S4. Further determine the uncertain areas and stitch them together with the structural features of the cloud and fog area to complete the cloud and fog complexity identification of the optical image.

[0005] Preferably, the process of S1 is as follows: S11. Calculate the grayscale mean of the optical image to be detected. Gray standard deviation Gray-scale histogram entropy and effective pixel ratio ; S12. Determine whether the optical image to be detected belongs to an extreme scene or a non-extreme scene based on the parameters calculated in S11; S13. If the optical image to be tested is an extreme scene, then reselect the optical image to be tested; if it is a non-extreme scene, then complete the confirmation of the extreme cloud and fog scene.

[0006] Preferably, the calculation process in S11 is as follows: ; ; ; ; in Indicates the image width. Indicates the image height. This represents the input optical image. Represents the horizontal coordinate of the pixel value. Represents the vertical coordinate of the pixel value. Represents the number of gray levels. Represents grayscale value The probability of occurrence This indicates the number of pixels whose grayscale values ​​are within the valid range.

[0007] Preferably, the process for determining extreme scenarios in S12 is as follows: like and If the optical image to be detected is determined to be a scene with full cloud coverage, the full cloud determination result will be output. in, Indicates the total cloud temperature threshold. Indicates the low variance threshold; like , and If the optical image to be detected is determined to be a cloudless scene, the cloudless determination result is output. in, The cloudless temperature threshold The minimum entropy threshold; like and If the optical image to be detected is determined to be a scene with strong noise interference, an invalid determination result will be output. in, For high variance threshold, Minimum percentage of effective pixels; like If the optical image to be detected is determined to be an image failure scene, an invalid determination result will be output. in, The threshold for determining image failure; If the above conditions are not met, the optical image to be detected is determined to be a non-extreme scene.

[0008] Preferably, the process of S2 is as follows: S21. Infrared image of the optical image to be detected. Perform top-hat transformation to obtain an enhanced image of the cloud / fog area. The process is as follows: ; in Representing an image Regarding structural elements Opening operation, It is a disk-shaped structural element with a radius of . ; S22, Through Structural elements of different scales To each Perform morphological filtering to obtain Amplitude response image , of which A response image as follows: ; in This represents the expansion operation. This represents the erosion operation. For the first There are structuring elements, with a scale parameter of . ; S23. Weighted fusion of responses at each scale is performed to obtain a fused response image. The process is as follows: ; ; in Indicates the first Weighting coefficients for each scale; S24. Based on the fused response image, the cloud and fog pixel region of the optical image to be detected is divided into deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region.

[0009] Preferably, the discrimination process of S24 is as follows: like If it is determined to be a definitive fog or cloud area, then it is considered a definitive fog or cloud area. This represents a preset deterministic cloud and fog threshold. like If it is determined to be a non-cloudy / foggy area, then it is considered a deterministic non-cloudy / foggy area. This indicates a preset deterministic non-cloud / fog threshold. like If it is, then it is determined to be a non-determined region.

[0010] Preferably, the process of S3 is as follows: S31. Perform morphological closing operation on the pixel-level cloud mask formed by the deterministic cloud region to connect the deterministic cloud regions in the optical image to be detected. S32. After the connection is completed, perform morphological opening operations to eliminate isolated noise points; After S33 and S32 are completed, the connected component labeling algorithm is used to remove components with areas smaller than [the specified area]. Isolated cloud areas and filling areas smaller than The internal pores were used to obtain the structural features of the cloud and fog area, forming a preliminary pixel-level cloud and fog mask. This represents the minimum effective cloud / fog area threshold. This is the threshold for the maximum area to fill the hole.

[0011] Preferably, the process of S4 is as follows: S41. Using pixels in the undetermined region as the basis for judgment. Centered on, construct a structure of size Local window Calculate the average gray level within the window. and standard deviation The process is as follows: ; ; in This indicates the default window standard; Indicates the location is The grayscale value at that location; S42. Calculate pixel-level adaptive thresholds for non-deterministic regions based on local statistical characteristics. The process is as follows: ; ; in This represents the local standard deviation adjustment factor. This represents the Laplace edge penalty coefficient. Indicates the image in The Laplace operator response at the location; S43. Combine the grayscale values ​​of the non-deterministic region with... The process of comparing, determining, and updating the cloud and fog status is as follows: If the grayscale value is less than If the region is not defined, it is determined to be a cloud or fog pixel region; otherwise, it is determined to be a non-cloud or fog pixel region. S44. Confirm the updated cloud and fog status to complete the cloud and fog complexity identification in the optical image. The process is as follows: If the proportion of non-cloud pixels in the eight neighborhoods of a cloud pixel exceeds If so, the pixel will be reclassified as a non-cloud pixel; If the proportion of cloud pixels within the eight neighborhoods of a non-cloud pixel exceeds If so, the pixel will be reclassified as a cloud pixel; in This indicates the preset threshold for determining isolated cloud pixels. This indicates the preset threshold for determining isolated non-cloud pixels.

[0012] Preferably, the method for identifying the complexity of clouds and fog in optical images is implemented through an optical image cloud and fog complexity identification system, including an extreme scene judgment module, a morphological judgment module, and a threshold segmentation supplementary judgment module; The extreme scene judgment module is used to perform global statistical feature analysis on the input infrared image to determine whether the current image belongs to an extreme scene; The morphological determination module uses multi-scale morphological filtering to extract structural features of cloud and fog regions, generate cloud and fog pixel region masks, and determine uncertain cloud and fog regions. The threshold segmentation supplementary judgment module combines the neighborhood grayscale distribution characteristics and uses an adaptive threshold segmentation algorithm to supplement the judgment of uncertain cloud and fog areas.

[0013] Preferably, the morphological determination module includes a background suppression model, a multi-structural element filtering model, a feature fusion and preliminary judgment model, and a cloud and fog area merging model; The threshold segmentation supplementary judgment module includes a neighborhood statistical feature extraction model, a dynamic threshold calculation model, and a supplementary discrimination model.

[0014] Therefore, the present invention adopts the above-mentioned method for identifying cloud and fog complexity in optical images. By designing a three-layer judgment structure of extreme scene judgment, morphological judgment, and threshold segmentation supplementary judgment, it can achieve pixel-level identification of cloud and fog complexity in optical images under complex scenes, thus supporting the improvement of optical image perception capabilities.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the optical image cloud and fog complexity identification method of the present invention; Figure 2 This is a model diagram of an optical image cloud and fog complexity identification method according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Specific model specifications need to be selected and determined according to the actual specifications of the device, etc. The specific selection calculation method adopts existing technology in the art, and therefore will not be described in detail.

[0018] Example like Figure 1 and Figure 2 As shown, the present invention provides a method for identifying the complexity of cloud and fog in optical images, comprising the following steps: S1. Confirm the extreme cloud and fog scene of the optical image to be tested; The extreme scene judgment module performs global statistical feature analysis on the entire input image to determine whether the current state belongs to an extreme scene; S11. Calculate the grayscale mean of the optical image to be detected. Gray standard deviation Gray-scale histogram entropy and effective pixel ratio ; ; ; ; ; in Indicates the image width. Indicates the image height. This represents the input optical image. Represents the horizontal coordinate of the pixel value. Represents the vertical coordinate of the pixel value. Represents the number of gray levels. Represents grayscale value The probability of occurrence This indicates the number of pixels whose grayscale values ​​are within the valid range.

[0019] S12. Determine whether the optical image to be detected belongs to an extreme scene or a non-extreme scene based on the parameters calculated in S11; like and If the optical image to be detected is determined to be a scene with full cloud coverage, the full cloud determination result will be output. in, Indicates the total cloud temperature threshold. Indicates the low variance threshold; like , and If the optical image to be detected is determined to be a cloudless scene, the cloudless determination result is output. in, The cloudless temperature threshold The minimum entropy threshold; like and If the optical image to be detected is determined to be a scene with strong noise interference, an invalid determination result will be output. in, For high variance threshold, Minimum percentage of effective pixels; like If the optical image to be detected is determined to be an image failure scene, an invalid determination result will be output. in, The threshold for determining image failure; If the above conditions are not met, the optical image to be detected is determined to be a non-extreme scene.

[0020] S13. If the optical image to be tested is an extreme scene, then reselect the optical image to be tested; if it is a non-extreme scene, then complete the confirmation of the extreme cloud and fog scene.

[0021] S2. After completing S1, determine the cloud and fog pixel region of the optical image to be detected, and complete the determination of deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region. The morphological judgment module performs multi-scale morphological filtering on infrared images of non-extreme scenes, extracts structural features of cloud and fog areas, forms a preliminary pixel-level cloud and fog mask, and judges deterministic and indeterminate cloud and fog areas. It mainly includes four models: background suppression, multi-structural element filtering, feature fusion and preliminary judgment, and cloud and fog area merging. S21, Background suppression model: Infrared image of the optical image to be detected. Perform top-hat transformation to obtain an enhanced image of the cloud / fog area. The process is as follows: ; in Representing an image Regarding structural elements Opening operation, It is a disk-shaped structural element with a radius of . ; S22, Multi-structural element filtering model through Structural elements of different scales To each Perform morphological filtering to obtain Amplitude response image , of which A response image as follows: ; in This represents the expansion operation. This represents the erosion operation. For the first There are structuring elements, with a scale parameter of . ; S23. The feature fusion and preliminary judgment model performs weighted fusion of responses at each scale to obtain a fused response image. The process is as follows: ; ; in Indicates the first Weighting coefficients for each scale; S24. Based on the fused response image, the cloud and fog pixel region of the optical image to be detected is divided into deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region.

[0022] like If it is determined to be a definitive fog or cloud area, then it is considered a definitive fog or cloud area. This represents a preset deterministic cloud and fog threshold. like If it is determined to be a non-cloudy / foggy area, then it is considered a deterministic non-cloudy / foggy This indicates a preset deterministic non-cloud / fog threshold. like If it is, then it is determined to be a non-determined region.

[0023] S3. Connect the deterministic cloud and fog regions in the optical image to be detected to obtain the structural features of the cloud and fog regions and form a preliminary pixel-level cloud and fog mask. S31. The cloud and fog merging model performs morphological closing operations on the pixel-level cloud and fog mask formed by the deterministic cloud and fog region to connect the deterministic cloud and fog regions in the optical image to be detected. S32. After the connection is completed, perform morphological opening operations to eliminate isolated noise points; After S33 and S32 are completed, the connected component labeling algorithm is used to remove components with areas smaller than [the specified area]. Isolated cloud areas and filling areas smaller than The internal pores were used to obtain the structural features of the cloud and fog area, forming a preliminary pixel-level cloud and fog mask. This represents the minimum effective cloud / fog area threshold. This is the threshold for the maximum area to fill the hole.

[0024] S4. Further determine the uncertain areas and stitch them together with the structural features of the cloud and fog area to complete the cloud and fog complexity identification of the optical image.

[0025] The threshold segmentation supplementary judgment module targets uncertain regions and combines the gray-scale distribution characteristics of the neighborhood to perform supplementary judgment using an adaptive threshold segmentation algorithm, outputting the final pixel-level cloud and fog state identification results. It mainly includes three models: neighborhood statistical feature extraction, dynamic threshold calculation, and supplementary judgment. S41, Neighborhood statistical feature extraction model uses pixels to be judged in non-deterministic regions. Centered on, construct a structure of size Local window Calculate the average gray level within the window. and standard deviation The process is as follows: ; ; in This indicates the preset window standard, and can take values ​​of 1, 2, or 3. Indicates the location is The grayscale value at that location; S42. The dynamic threshold calculation model calculates pixel-level adaptive thresholds for non-deterministic regions based on local statistical characteristics. The process is as follows: ; ; in This represents the local standard deviation adjustment factor. This represents the Laplace edge penalty coefficient. Indicates the image in The Laplace operator response at the location; S43. The supplementary discriminant model compares the gray values ​​of uncertain regions with... The process of comparing, determining, and updating the cloud and fog status is as follows: If the grayscale value is less than If the region is not defined, it is determined to be a cloud or fog pixel region; otherwise, it is determined to be a non-cloud or fog pixel region. S44. Confirm the updated cloud and fog status to complete the cloud and fog complexity identification in the optical image. The process is as follows: If the proportion of non-cloud pixels in the eight neighborhoods of a cloud pixel exceeds If so, the pixel will be reclassified as a non-cloud pixel; If the proportion of cloud pixels within the eight neighborhoods of a non-cloud pixel exceeds If so, the pixel will be reclassified as a cloud pixel; in This indicates the preset threshold for determining isolated cloud pixels (typically 0.75). This represents the preset threshold for determining isolated non-cloud pixels (typically 0.75).

[0026] An optical image cloud and fog complexity identification method is implemented through an optical image cloud and fog complexity identification system, including an extreme scene judgment module, a morphological judgment module, and a threshold segmentation supplementary judgment module; The extreme scene judgment module is used to perform global statistical feature analysis on the input infrared image to determine whether the current image belongs to an extreme scene; The morphological determination module uses multi-scale morphological filtering to extract structural features of cloud and fog regions, generate cloud and fog pixel region masks, and determine uncertain cloud and fog regions. The threshold segmentation supplementary judgment module combines the neighborhood grayscale distribution characteristics and uses an adaptive threshold segmentation algorithm to supplement the judgment of uncertain cloud and fog areas.

[0027] The morphological determination module includes a background suppression model, a multi-structural element filtering model, a feature fusion and preliminary judgment model, and a cloud and fog area merging model; The threshold segmentation supplementary judgment module includes a neighborhood statistical feature extraction model, a dynamic threshold calculation model, and a supplementary discrimination model.

[0028] Therefore, this invention employs an optical image cloud and fog complexity identification method. By designing a three-layer judgment structure consisting of extreme scene judgment, morphological judgment, and threshold segmentation supplementary judgment, it can achieve pixel-level identification of cloud and fog complexity in optical images under complex scenes, thereby supporting the improvement of optical image perception capabilities.

[0029] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying the complexity of clouds and fog in optical images, characterized in that, Includes the following steps: S1. Confirm the extreme cloud and fog scene of the optical image to be tested; S2. After completing S1, determine the cloud and fog pixel region of the optical image to be detected, and complete the determination of deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region. S3. Connect the deterministic cloud and fog regions in the optical image to be detected to obtain the structural features of the cloud and fog regions and form a preliminary pixel-level cloud and fog mask. S4. Further determine the uncertain areas and stitch them together with the structural features of the cloud and fog area to complete the cloud and fog complexity identification of the optical image.

2. The method for identifying the complexity of cloud and fog in optical images according to claim 1, characterized in that, The process of S1 is as follows: S11. Calculate the grayscale mean of the optical image to be detected. Gray standard deviation Gray-scale histogram entropy and effective pixel ratio ; S12. Determine whether the optical image to be detected belongs to an extreme scene or a non-extreme scene based on the parameters calculated in S11; S13. If the optical image to be tested is an extreme scene, then reselect the optical image to be tested; if it is a non-extreme scene, then complete the confirmation of the extreme cloud and fog scene.

3. The method for identifying the complexity of cloud and fog in optical images according to claim 2, characterized in that, The calculation process in S11 is as follows: ； ； ； ； in Indicates the image width. Indicates the image height. This represents the input optical image. Represents the horizontal coordinate of the pixel value. Represents the vertical coordinate of the pixel value. Represents the number of gray levels. Represents grayscale value The probability of occurrence This indicates the number of pixels whose grayscale values ​​are within the valid range.

4. The method for identifying the complexity of cloud and fog in optical images according to claim 3, characterized in that, The process for determining extreme scenarios in S12 is as follows: like and If the optical image to be detected is determined to be a scene with full cloud coverage, the full cloud determination result will be output. in, Indicates the total cloud temperature threshold. Indicates the low variance threshold; like , and If the optical image to be detected is determined to be a cloudless scene, the cloudless determination result is output. in, The cloudless temperature threshold The minimum entropy threshold; like and If the optical image to be detected is determined to be a scene with strong noise interference, an invalid determination result will be output. in, For high variance threshold, Minimum percentage of effective pixels; like If the optical image to be detected is determined to be an image failure scene, an invalid determination result will be output. in, The threshold for determining image failure; If the above conditions are not met, the optical image to be detected is determined to be a non-extreme scene.

5. The method for identifying the complexity of cloud and fog in optical images according to claim 4, characterized in that, The process of S2 is as follows: S21. Infrared image of the optical image to be detected. Perform top-hat transformation to obtain an enhanced image of the cloud / fog area. The process is as follows: ； in Representing an image Regarding structural elements Opening operation, It is a disk-shaped structural element with a radius of . ; S22, Through Structural elements of different scales To each Perform morphological filtering to obtain Amplitude response image , of which A response image as follows: ； in This represents the expansion operation. This represents the erosion operation. For the first There are structuring elements, with a scale parameter of . ; S23. Weighted fusion of responses at each scale is performed to obtain a fused response image. The process is as follows: ； ； in Indicates the first Weighting coefficients for each scale; S24. Based on the fused response image, the cloud and fog pixel region of the optical image to be detected is divided into deterministic cloud and fog region, deterministic non-cloud and fog region and indeterminate region.

6. The method for identifying the complexity of cloud and fog in optical images according to claim 5, characterized in that, The discrimination process for S24 is as follows: like If it is determined to be a definitive fog or cloud area, then it is considered a definitive fog or cloud area. This represents a preset deterministic cloud and fog threshold. like If it is determined to be a non-cloudy / foggy area, then it is considered a deterministic non-cloudy / foggy This indicates a preset deterministic non-cloud / fog threshold. like If it is, then it is determined to be a non-determined region.

7. The method for identifying the complexity of cloud and fog in optical images according to claim 6, characterized in that, The process of S3 is as follows: S31. Perform morphological closing operation on the pixel-level cloud mask formed by the deterministic cloud region to connect the deterministic cloud regions in the optical image to be detected. S32. After the connection is completed, perform morphological opening operations to eliminate isolated noise points; After S33 and S32 are completed, the connected component labeling algorithm is used to remove components with areas smaller than [the specified area]. Isolated cloud areas and filling areas smaller than The internal pores were used to obtain the structural features of the cloud and fog area, forming a preliminary pixel-level cloud and fog mask. This represents the minimum effective cloud / fog area threshold. This is the threshold for the maximum area to fill the hole.

8. The method for identifying the complexity of cloud and fog in optical images according to claim 7, characterized in that, The process of S4 is as follows: S41. Using pixels in the undetermined region as the basis for judgment. Centered on, construct a structure of size Local window Calculate the average gray level within the window. and standard deviation The process is as follows: ； ； in This indicates the default window standard; Indicates the location is The grayscale value at that location; S42. Calculate pixel-level adaptive thresholds for non-deterministic regions based on local statistical characteristics. The process is as follows: ； ； in This represents the local standard deviation adjustment factor. This represents the Laplace edge penalty coefficient. Indicates the image in The Laplace operator response at the location; S43. Combine the grayscale values ​​of the non-deterministic region with... The process of comparing, determining, and updating the cloud and fog status is as follows: If the grayscale value is less than If the region is not defined, it is determined to be a cloud or fog pixel region; otherwise, it is determined to be a non-cloud or fog pixel region. S44. Confirm the updated cloud and fog status to complete the cloud and fog complexity identification in the optical image. The process is as follows: If the proportion of non-cloud pixels in the eight neighborhoods of a cloud pixel exceeds If so, the pixel will be reclassified as a non-cloud pixel; If the proportion of cloud pixels within the eight neighborhoods of a non-cloud pixel exceeds If so, the pixel will be reclassified as a cloud pixel; in This indicates the preset threshold for determining isolated cloud pixels. This indicates the preset threshold for determining isolated non-cloud pixels.

9. The method for identifying the complexity of cloud and fog in optical images according to claim 8, characterized in that: An optical image cloud and fog complexity identification method is implemented through an optical image cloud and fog complexity identification system, including an extreme scene judgment module, a morphological judgment module, and a threshold segmentation supplementary judgment module; The extreme scene judgment module is used to perform global statistical feature analysis on the input infrared image to determine whether the current image belongs to an extreme scene; The morphological determination module uses multi-scale morphological filtering to extract structural features of cloud and fog regions, generate cloud and fog pixel region masks, and determine uncertain cloud and fog regions. The threshold segmentation supplementary judgment module combines the neighborhood grayscale distribution characteristics and uses an adaptive threshold segmentation algorithm to supplement the judgment of uncertain cloud and fog areas.

10. The method for identifying the complexity of cloud and fog in optical images according to claim 9, characterized in that: The morphological determination module includes a background suppression model, a multi-structural element filtering model, a feature fusion and preliminary judgment model, and a cloud and fog area merging model; The threshold segmentation supplementary judgment module includes a neighborhood statistical feature extraction model, a dynamic threshold calculation model, and a supplementary discrimination model.

Images