High-precision sky cloud detection method, system and sensor
By constructing a light intensity reference benchmark and RGB ratio and gradient analysis, the system achieves automated and accurate differentiation of sky cloud cover detection, solving the problem of distinguishing between light changes and thin clouds/haze in existing technologies, and improving detection accuracy and applicability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING WEATHER MODIFICATION OFFICE
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-16
AI Technical Summary
Existing methods for detecting cloud cover are greatly affected by dynamic changes in illumination, making it difficult to adapt to complex sky scenes. They have a high false alarm rate, difficulty in distinguishing between thin clouds and haze, and lack standardized reference benchmarks and automated processing, resulting in insufficient detection accuracy and applicability.
By constructing a reference benchmark under illumination intensity and combining RGB ratios and gradient analysis, accurate differentiation of sky, clouds, thin clouds and haze areas is achieved. Triangular feature mapping and Sobel gradient algorithm are used to build an automated detection process, reducing reliance on manual labor.
It improves the accuracy and applicability of cloud cover detection, reduces false alarms, adapts to different lighting conditions, automates processing to reduce operational complexity, and adapts to complex sky scenarios.
Smart Images
Figure CN121921658B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cloud cover detection technology, specifically to a high-precision sky cloud cover detection method, system, and sensor. Background Technology
[0002] Cloud cover is a core fundamental parameter in fields such as meteorological observation, photovoltaic power prediction, aviation safety assurance, and climate research. Its detection accuracy directly affects the reliability and accuracy of various related applications. As the requirements for data accuracy in related fields continue to increase, traditional cloud cover detection methods can no longer meet the needs of practical applications, and there are many technical bottlenecks that need to be overcome.
[0003] Current methods for detecting cloud cover rely heavily on manual observation or simple image thresholding. Manual observation is affected by the subjective judgment of observers, resulting in low detection efficiency, large errors, and the inability to achieve continuous real-time monitoring. Simple thresholding methods typically use fixed RGB thresholds to divide the sky and cloud regions, without considering the impact of dynamic changes in light intensity on the RGB values of the sky and clouds. The RGB values of the same region fluctuate greatly under different lighting conditions, while the RGB values of the sky and clouds exhibit clustering characteristics under the same lighting conditions. Fixed thresholds cannot adapt to scenarios with dynamic changes in lighting, which can easily lead to misjudgment of sky and cloud regions, thereby affecting the accuracy of cloud cover statistics.
[0004] Furthermore, existing methods struggle to effectively distinguish between thin clouds and haze areas. Both exhibit high brightness in the visible light spectrum, with similar RGB values. Traditional detection methods based solely on color features cannot accurately differentiate between the two, easily misclassifying haze areas as thin cloud areas or omitting thin cloud areas, leading to significant deviations in cloud cover statistics. Simultaneously, some detection methods lack standardized reference benchmarks, relying solely on discrete RGB value comparisons for area calibration. This lack of data support and verification processes results in insufficient stability and reliability of the detection results, making them unsuitable for complex sky scenarios such as clear skies, cloudy skies, and mixed haze.
[0005] Furthermore, existing detection methods often lack layered detection logic and a systematic approach to classifying the sky, clouds, thin clouds, and haze. They either ignore the identification of haze areas, fail to design specific judgment mechanisms for the texture and color characteristics of thin clouds, or fail to utilize texture features such as gradients to help distinguish easily confused areas, thus limiting the applicability of the detection. At the same time, some methods are highly dependent on the experience of operators, lack quantifiable judgment standards and automated processing procedures, have poor engineering feasibility, and are difficult to meet the needs of high-precision, real-time cloud cover monitoring.
[0006] Therefore, developing a high-precision sky cloud cover detection method that can adapt to dynamic changes in illumination, accurately distinguish various easily confused areas, construct a reliable reference benchmark, reduce reliance on manual labor, and has high applicability has become an urgent technical problem to be solved. Summary of the Invention
[0007] To address the shortcomings of existing technologies, this invention provides a high-precision sky cloud cover detection method, system, and sensor, which solves the problem that some detection methods do not establish a standardized reference benchmark and only perform regional calibration by comparing discrete RGB values, lacking data support and verification.
[0008] To achieve the above objectives, the present invention provides a high-precision sky cloud cover detection method, comprising the following steps:
[0009] Step 1: Identify the RGB data associated with the same illumination intensity in the sky and cloud regions from historical imaging data, and generate reference benchmarks associated with the sky and cloud regions respectively based on the identified RGB data. The specific method is as follows:
[0010] Extract historical images belonging to the same illumination intensity from historical imaging data, then identify sky images of the sky region from the historical imaging data, and simultaneously extract cloud images of the cloud region.
[0011] The system identifies the RGB values associated with individual image points in the sky image and generates a set of triangles. Simultaneously, the center point of the triangle is recorded as the initial point, and the corner points of the triangle are recorded as the end points. Three sets of baselines are generated, and the three sets of baselines are sequentially associated with R, G, or B values. Based on the specific type of numerical value associated with the corresponding baseline, a measurement length associated with each unit length of the corresponding baseline is assigned. According to the different RGB values associated with different image points in the sky image, the measurement points associated with the corresponding image points are locked on the corresponding baselines. The three sets of measurement points are connected and processed. The triangle obtained by the connection is recorded as the feature polygon associated with the corresponding image point. Then, the feature polygons associated with different image points are sequentially confirmed, and the area included between the confirmed feature polygons is recorded as the reference area. The confirmed reference area is recorded as the reference baseline for sky imaging under the current illumination intensity.
[0012] Using a confirmation method that identifies the same reference benchmark from sky imaging, the RGB values associated with different image points in cloud imaging are confirmed, corresponding triangles are constructed simultaneously, the feature polygons associated with different image points are locked, and the internal regions included by several sets of feature polygons are recorded as the reference benchmarks associated with the corresponding cloud imaging.
[0013] Step 2: Acquire the regional image of the corresponding sky area, determine the RGB values associated with different image points in the regional image, extract the associated reference benchmark based on the illumination intensity associated at the acquisition time, and calibrate the associated sky and cloud areas within the regional image based on the reference benchmark. The specific method is as follows:
[0014] Based on the photosensitive sensor installed in the all-sky imager, the illumination intensity associated with the acquired area image is confirmed, and based on the confirmed illumination intensity, the associated reference benchmark is extracted.
[0015] The reference reference for sky imaging is designated as the sky reference, and the reference reference for cloud imaging is designated as the cloud reference. The RGB values associated with a single image point are identified within the regional image. Within the constructed triangle, the different measurement points associated with the corresponding RGB values on different measurement lines are identified. Several sets of measurement points are connected to identify the triangle to be verified associated with the corresponding image point. The triangle to be verified is compared with the sky reference or the cloud reference. If the triangle to be verified is ∈ sky reference, the corresponding image point is designated as the sky image point. If the triangle to be verified is ∈ cloud reference, the corresponding image point is designated as the cloud image point.
[0016] Based on the sky and cloud image points marked sequentially within the regional image, the existing sky and cloud areas within the regional image are marked sequentially, and other unmarked areas are marked as unmarked areas;
[0017] Step 3: For other areas within the regional image that are not designated as sky or cloud areas, based on the RGB value features associated with different image points within these other areas, identify whether the corresponding image points belong to thin cloud image points and mark the thin cloud areas accordingly. The specific method is as follows:
[0018] Identify the RGB values associated with different image points in other regions and label them sequentially as R. i G i And B i Where i represents different image points, the two sets of ratios associated with a single image point are identified, namely (R... i / B i ) and (G i / B i The two confirmed ratios are then compared to confirm the difference, and it is determined whether the confirmed difference satisfies the following condition: difference ≤ Y1, where Y1 is a preset value. If the condition is met, the corresponding image point is recorded as a point to be associated. If the condition is not met, no marking is performed.
[0019] Identify B from the marked undetermined association points i R i G i The numerical difference will satisfy: |B i -R i |≤Y2、|G i -R i |≤Y2 and|G i -B iUndetermined associated points with |≤Y2 are denoted as thin cloud image points, where Y2 is a preset value. Thin cloud image points existing in other areas are marked sequentially, and thin cloud areas are marked in other areas based on the marked thin cloud image points.
[0020] Step 4: For the remaining areas not identified as thin cloud areas, perform grayscale processing on these remaining areas to confirm the grayscale imagery. Then, verify the gradient data associated with different image points within the grayscale imagery to identify whether the remaining areas belong to haze areas. The specific method is as follows:
[0021] Based on the remaining areas associated with other areas, the remaining areas are processed into grayscale to confirm the grayscale image associated with the remaining areas.
[0022] The Sobel algorithm is used to calibrate the gradient data associated with different image points within a grayscale image. First, the vertical gradient associated with a given image point is identified, followed by the horizontal gradient. The comprehensive gradient associated with the corresponding image point is confirmed, and the standard deviation of the comprehensive gradient associated with several image points is processed. The standard gradient associated with several comprehensive gradients is confirmed, and it is identified whether the confirmed standard gradient satisfies: standard gradient ≤ Y3. If it satisfies, the remaining area is labeled as haze area. If it does not satisfy, no labeling is performed. Y3 is a preset value.
[0023] Based on the area ratio of marked cloud regions and thin cloud regions within the regional image, cloud cover ratio is generated and directly output.
[0024] Preferably, a high-precision sky cloud cover detection system includes:
[0025] At the reference generation end, based on historical imaging data associated with different light intensities, the RGB data associated with the calibrated sky and cloud regions are identified, and reference benchmarks associated with the sky and cloud regions are generated based on the identified RGB data.
[0026] The regional calibration end acquires regional images of the corresponding sky area through an all-sky imager, determines the RGB values associated with different image points in the regional image, extracts the associated reference benchmark based on the light intensity associated at the time of acquisition, and calibrates the associated sky area and cloud area in the regional image based on the reference benchmark.
[0027] The thin cloud marking processing end identifies whether the corresponding image points belong to thin cloud image points and marks the thin cloud area for other areas in the regional image that are not marked as sky areas and cloud areas, based on the RGB value features associated with different image points in other areas.
[0028] The haze area marking end performs grayscale processing on the remaining areas in other areas that have not been marked as thin cloud areas to confirm the grayscale area image, and confirms the gradient data associated with different image points in the grayscale area image to identify whether the remaining area belongs to the haze area.
[0029] Preferably, the high-precision sky cloud cover detection sensor has a detection camera and a photosensitive sensor at its front end. The detection camera is used to detect the regional image of a designated area, and the photosensitive sensor is used to detect the light intensity in real time.
[0030] This invention provides a high-precision sky cloud cover detection method, system, and sensor. Compared with existing technologies, it has the following advantages:
[0031] This invention quickly identifies sky and cloud regions by comparing a reference benchmark with the triangle to be verified, reducing misjudgments of basic regions. The second step, for regions with similar RGB features, uses a dual-ratio difference and numerical difference assessment to accurately filter thin cloud image points, solving the problem of distinguishing thin clouds from other regions. The third step, leveraging grayscale processing and Sobel gradient analysis, determines haze regions by comprehensively considering the gradient standard deviation, effectively separating the textural differences between thin clouds and haze, avoiding cloud cover statistical bias caused by confusion between the two, and ultimately significantly improving the recognition accuracy of cloud and thin cloud regions, ensuring the accuracy of cloud cover percentage calculation.
[0032] By matching light intensity to corresponding reference benchmarks, it adapts to different lighting scenarios under different time periods and weather conditions. It also designs exclusive judgment logic for the differences in RGB and texture features of thin clouds and haze, adapting to complex sky scenarios such as clear skies, cloudy skies, and haze. At the same time, it builds reference benchmarks based on historical data, which has a certain degree of self-adaptation capability. It can optimize the benchmark accuracy by iteratively updating historical data, without the need for frequent adjustments to preset parameters, reducing the reliance on operator experience and enhancing the versatility and robustness of the method in practical engineering applications.
[0033] The entire detection process is clear and logically coherent. The techniques used, such as triangle feature mapping, RGB ratio calculation, Sobel gradient analysis, and grayscale processing, are all mature image processing algorithms. They do not require complex hardware support and can directly achieve data acquisition and detection calculations based on all-sky imagers and photosensitive sensors. The engineering implementation cost is low and easy to implement. Attached Figure Description
[0034] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation
[0035] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0036] First Embodiment
[0037] Please see Figure 1 This application provides a high-precision sky cloud cover detection method, including the following steps:
[0038] Step 1: Based on historical imaging data associated with different light intensities, identify the RGB data associated with the calibrated sky and cloud regions, and generate reference benchmarks associated with the sky and cloud regions based on the identified RGB data. Specifically, under different light intensities, the sky and clouds exhibit different RGB values, but under the same light intensities, the RGB data generated by the sky and clouds are relatively clustered, which can effectively perform comprehensive verification based on historical data and generate reference benchmarks associated with the corresponding regions.
[0039] The specific method for generating the reference benchmarks for the sky and cloud regions is as follows:
[0040] Extract historical images belonging to the same illumination intensity from historical imaging data, then identify sky images of the sky region from the historical imaging data, and simultaneously extract cloud images of the cloud region.
[0041] The system identifies the RGB values associated with individual image points in the sky image and generates a set of triangles. Simultaneously, the center point of the triangle is recorded as the initial point, and the corner points of the triangle are recorded as the end points. Three sets of baselines are generated, and the three sets of baselines are sequentially associated with R, G, or B values. Based on the specific type of numerical value associated with the corresponding baseline, a measurement length associated with each unit length of the corresponding baseline is assigned. According to the different RGB values associated with different image points in the sky image, the measurement points associated with the corresponding image points are locked on the corresponding baselines. The three sets of measurement points are connected and processed. The triangle obtained by the connection is recorded as the feature polygon associated with the corresponding image point. Then, the feature polygons associated with different image points are sequentially confirmed, and the area included between the confirmed feature polygons is recorded as the reference area. The confirmed reference area is recorded as the reference baseline for sky imaging under the current illumination intensity.
[0042] Using a confirmation method that identifies the same reference benchmark from sky imaging, the RGB values associated with different image points in cloud imaging are confirmed, corresponding triangles are constructed simultaneously, the feature polygons associated with different image points are locked, and the internal regions included by several sets of feature polygons are recorded as the reference benchmarks associated with the corresponding cloud imaging.
[0043] Specifically, in historical imaging data under the same light intensity, there are sky images and cloud images. In the sky image, the corresponding image points are associated with corresponding RGB values (R value, G value and B value). Based on the measurement points that exist on the measurement line with different values, triangles associated with multiple measurement points can be effectively generated. Based on the specific areas included between multiple triangles, the corresponding reference benchmark can be effectively locked, which facilitates parameter verification when performing numerical comparisons later.
[0044] Furthermore, the historical imaging data already included pre-marked sky or cloud images;
[0045] Step 2: Acquire the regional image of the corresponding sky area, determine the RGB values associated with different image points in the regional image, extract the associated reference benchmark based on the light intensity associated at the time of acquisition, and calibrate the associated sky area and cloud area in the regional image based on the reference benchmark. Specifically, in the corresponding regional image, based on the RGB values associated with the corresponding image points and the constructed reference benchmark, the associated sky area and cloud area in the regional image can be effectively calibrated sequentially.
[0046] The specific method for marking the sky and cloud areas is as follows:
[0047] Based on the photosensitive sensor installed in the all-sky imager, the illumination intensity associated with the acquired area image is confirmed, and based on the confirmed illumination intensity, the associated reference benchmark is extracted.
[0048] The reference reference for sky imaging is designated as the sky reference, and the reference reference for cloud imaging is designated as the cloud reference. The RGB values associated with a single image point are identified within the regional image. Within the constructed triangle, the different measurement points associated with the corresponding RGB values on different measurement lines are identified. Several sets of measurement points are connected to identify the triangle to be verified associated with the corresponding image point. The triangle to be verified is compared with the sky reference or the cloud reference. If the triangle to be verified is ∈ Sky reference (that is, the corresponding triangle is located within the area of the corresponding reference), the corresponding image point is designated as a sky image point. If the triangle to be verified is ∈ Cloud reference, the corresponding image point is designated as a cloud image point.
[0049] Based on the sky and cloud image points marked sequentially within the regional image, the existing sky and cloud regions within the regional image are marked sequentially, and other unmarked regions are marked as undetermined regions. Undetermined regions are those whose RGB values associated with the corresponding image points do not belong to either the sky or cloud reference. Such regions may belong to thin cloud regions or haze regions, because the RGB value characteristics associated with these two regions are quite similar. Therefore, further analysis and detection processes are needed to effectively divide the corresponding thin cloud regions or haze regions.
[0050] Step 3: For other areas within the regional image that are not labeled as sky or cloud areas, based on the RGB value features associated with different image points in other areas, identify whether the corresponding image points belong to thin cloud image points and mark the thin cloud areas. Specifically, in other areas, thin cloud areas and haze areas have different image point performance characteristics. In thin cloud areas, the R, G, and B values are relatively similar. In haze areas, the ratios of R and B, as well as G and B, are relatively similar, but the B value is slightly higher than G and R, which is quite different from the thin cloud areas.
[0051] The specific method for marking thin cloud areas in other regions is as follows:
[0052] Identify the RGB values associated with different image points in other regions and label them sequentially as R. i G i And B i Where i represents different image points, the two sets of ratios associated with a single image point are identified, namely (R... i / B i ) and (G i / B i The two confirmed ratios are then compared to confirm the difference, and it is determined whether the confirmed difference satisfies the following condition: difference ≤ Y1, where Y1 is a preset value, and its specific value is determined by the operator based on experience, generally taking the value of 2. If the condition is met, the corresponding image point is recorded as a pending association point; if the condition is not met, no marking is made.
[0053] Identify B from the marked undetermined association points i R i G i The numerical difference will satisfy: |B i -R i |≤Y2、|G i -R i |≤Y2 and|G i -B iThe undetermined associated points with |≤Y2 are recorded as thin cloud image points, where Y2 is a preset value. Its specific value is determined by the operator based on experience, and is generally 10. The thin cloud image points existing in other areas are marked in sequence, and the thin cloud areas are marked in other areas based on the marked thin cloud image points.
[0054] Within the thin cloud area, there are associated thin cloud points. The RGB values associated with these thin cloud points are quite similar. Therefore, based on the preset values, the thin cloud points associated with the thin cloud area can be effectively identified. Based on the thin cloud image points identified in sequence, the thin cloud area is marked sequentially, and the amount of cloud cover in the sky is effectively calibrated.
[0055] Step 4: For the remaining areas not identified as thin cloud areas, perform grayscale processing on these remaining areas to confirm the grayscale imagery. Then, verify the gradient data associated with different image points within the grayscale imagery to identify whether the remaining areas belong to haze areas.
[0056] Based on the remaining regions associated with other regions, the remaining regions are converted to grayscale to identify the associated grayscale image regions (different weights are assigned to RGB values to identify the associated grayscale values of corresponding image points; the grayscale value = 0.114 × R). i +0.587×G i +0.299×B i );
[0057] The Sobel algorithm is used to calibrate the gradient data associated with different image points within a grayscale image. First, the vertical gradient associated with a given image point is identified, followed by the horizontal gradient. Confirm the integrated gradient associated with the corresponding image point, and process the standard deviation of the integrated gradient associated with several image points to confirm the standard gradient associated with several integrated gradients. Then, identify whether the confirmed standard gradient satisfies the following condition: standard gradient ≤ Y3. If it satisfies the condition, the remaining area is labeled as a haze area. If it does not satisfy the condition, no labeling is performed. Y3 is a preset value, and its specific value is determined by the operator based on experience. The value of Y3 is between 8 and 15.
[0058] Based on the area ratio of marked cloud regions and thin cloud regions within the regional image, cloud cover ratio is generated and directly output.
[0059] Second Embodiment
[0060] A high-precision sky cloud cover detection system includes:
[0061] At the reference generation end, based on historical imaging data associated with different light intensities, the RGB data associated with the calibrated sky and cloud regions are identified, and reference benchmarks associated with the sky and cloud regions are generated based on the identified RGB data.
[0062] The regional calibration end acquires regional images of the corresponding sky area through an all-sky imager, determines the RGB values associated with different image points in the regional image, extracts the associated reference benchmark based on the light intensity associated at the time of acquisition, and calibrates the associated sky area and cloud area in the regional image based on the reference benchmark.
[0063] The thin cloud marking processing unit identifies whether other areas within the regional image that are not marked as sky or cloud areas belong to thin cloud image points and marks thin cloud areas based on the RGB value characteristics associated with different image points in other areas.
[0064] The haze area marking end performs grayscale processing on the remaining areas in other areas that have not been marked as thin cloud areas to confirm the grayscale area image, and confirms the gradient data associated with different image points in the grayscale area image to identify whether the remaining area belongs to the haze area.
[0065] Third Embodiment
[0066] The high-precision sky cloud cover detection system has a sensor at its front end, which is equipped with a detection camera and a photosensitive sensor. The detection camera is used to detect the regional image of a designated area, and the photosensitive sensor is used to detect the light intensity in real time.
[0067] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.
[0068] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. 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 be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A high-precision method for detecting cloud cover, characterized in that, Includes the following steps: Step 1: Identify the RGB data associated with the same illumination intensity in the sky and cloud regions from historical imaging data, and generate reference benchmarks associated with the sky and cloud regions respectively based on the identified RGB data. The specific method is as follows: Extract historical images belonging to the same illumination intensity from historical imaging data, then identify sky images of the sky region from the historical imaging data, and simultaneously extract cloud images of the cloud region. The system identifies the RGB values associated with individual image points in the sky image and generates a set of triangles. Simultaneously, the center point of the triangle is recorded as the initial point, and the corner points of the triangle are recorded as the end points. Three sets of baselines are generated, and the three sets of baselines are sequentially associated with R, G, or B values. Based on the specific type of numerical value associated with the corresponding baseline, a measurement length associated with each unit length of the corresponding baseline is assigned. According to the different RGB values associated with different image points in the sky image, the measurement points associated with the corresponding image points are locked on the corresponding baselines. The three sets of measurement points are connected and processed. The triangle obtained by the connection is recorded as the feature polygon associated with the corresponding image point. Then, the feature polygons associated with different image points are sequentially confirmed, and the area included between the confirmed feature polygons is recorded as the reference area. The confirmed reference area is recorded as the reference baseline for sky imaging under the current illumination intensity. Using a confirmation method that identifies the same reference benchmark from sky imaging, the RGB values associated with different image points in cloud imaging are confirmed, corresponding triangles are constructed simultaneously, the feature polygons associated with different image points are locked, and the internal regions included by several sets of feature polygons are recorded as the reference benchmarks associated with the corresponding cloud imaging. Step 2: Obtain the regional image of the corresponding sky area, determine the RGB values associated with different image points in the regional image, extract the associated reference benchmark based on the light intensity associated at the time of acquisition, and calibrate the associated sky area and cloud area in the regional image based on the reference benchmark. Step 3: For other areas in the regional image that are not marked as sky or cloud areas, based on the RGB value features associated with different image points in other areas, identify whether the corresponding image points belong to thin cloud image points and mark the thin cloud areas. Step 4: For the remaining areas in other regions that have not been identified as thin cloud areas, perform grayscale processing on the remaining areas to confirm the grayscale image, and confirm the gradient data associated with different image points in the grayscale image to identify whether the remaining areas belong to haze areas.
2. The high-precision sky cloud cover detection method according to claim 1, characterized in that, In step two, the specific method for marking the sky area and the cloud area is as follows: Based on the photosensitive sensor installed in the all-sky imager, the illumination intensity associated with the acquired area image is confirmed, and based on the confirmed illumination intensity, the associated reference benchmark is extracted. The reference reference for sky imaging is designated as the sky reference, and the reference reference for cloud imaging is designated as the cloud reference. The RGB values associated with a single image point are identified within the regional image. Within the constructed triangle, the different measurement points associated with the corresponding RGB values on different measurement lines are identified. Several sets of measurement points are connected to identify the triangle to be verified associated with the corresponding image point. The triangle to be verified is compared with the sky reference or the cloud reference. If the triangle to be verified is ∈ sky reference, the corresponding image point is designated as the sky image point. If the triangle to be verified is ∈ cloud reference, the corresponding image point is designated as the cloud image point. Based on the sky and cloud image points marked sequentially within the regional image, the existing sky and cloud areas within the regional image are marked sequentially, and other unmarked areas are marked as unmarked areas.
3. The high-precision sky cloud cover detection method according to claim 1, characterized in that, In step three, the specific method for marking thin cloud areas in other regions is as follows: Identify the RGB values associated with different image points in other regions and label them sequentially as R. i G i And B i Where i represents different image points, the two sets of ratios associated with a single image point are identified, namely (R... i / B i ) and (G i / B i The two confirmed ratios are then compared to confirm the difference, and it is determined whether the confirmed difference satisfies the following condition: difference ≤ Y1, where Y1 is a preset value. If the condition is met, the corresponding image point is recorded as a point to be associated. Identify B from the marked undetermined association points i R i G i The numerical difference will satisfy: |B i -R i |≤Y2、|G i -R i |≤Y2 and|G i -B i Points of unknown association with |≤Y2 are denoted as thin cloud image points, where Y2 is a preset value. Thin cloud image points existing in other areas are marked sequentially, and thin cloud areas are marked in other areas based on the marked thin cloud image points.
4. The high-precision sky cloud cover detection method according to claim 3, characterized in that, If the difference does not satisfy the condition: difference ≤ Y1, then no marking is performed.
5. The high-precision sky cloud cover detection method according to claim 1, characterized in that, In step four, the specific method for identifying whether the remaining area belongs to the haze area is as follows: Based on the remaining areas associated with other areas, the remaining areas are processed into grayscale to confirm the grayscale image associated with the remaining areas. The Sobel algorithm is used to calibrate the gradient data associated with different image points within a grayscale image. First, the vertical gradient associated with a given image point is identified, followed by the horizontal gradient. The comprehensive gradient associated with the corresponding image point is confirmed, and the standard deviation of the comprehensive gradient associated with several image points is processed. The standard gradient associated with several comprehensive gradients is confirmed, and it is identified whether the confirmed standard gradient satisfies: standard gradient ≤ Y3. If it satisfies, the remaining area is labeled as haze area. If it does not satisfy, no labeling is performed. Y3 is a preset value. Based on the area ratio of marked cloud regions and thin cloud regions within the regional image, cloud cover ratio is generated and directly output.
6. A high-precision sky cloud cover detection system, wherein the system operates according to the high-precision sky cloud cover detection method according to any one of claims 1-5, characterized in that, include: At the reference generation end, based on historical imaging data associated with different light intensities, the RGB data associated with the calibrated sky and cloud regions are identified, and reference benchmarks associated with the sky and cloud regions are generated based on the identified RGB data. The regional calibration end acquires regional images of the corresponding sky area through an all-sky imager, determines the RGB values associated with different image points in the regional image, extracts the associated reference benchmark based on the light intensity associated at the time of acquisition, and calibrates the associated sky area and cloud area in the regional image based on the reference benchmark. The thin cloud marking processing end identifies whether the corresponding image points belong to thin cloud image points and marks the thin cloud area for other areas in the regional image that are not marked as sky areas and cloud areas, based on the RGB value features associated with different image points in other areas. The haze area marking end performs grayscale processing on the remaining areas in other areas that have not been marked as thin cloud areas to confirm the grayscale area image, and confirms the gradient data associated with different image points in the grayscale area image to identify whether the remaining area belongs to the haze area.
7. A high-precision sky cloud cover detection sensor, wherein the sensor is applied in the high-precision sky cloud cover detection method according to any one of claims 1-5, characterized in that, Its sensor front end is equipped with a detection camera and a photosensitive sensor. The detection camera is used to detect the regional image of a designated area, and the photosensitive sensor is used to detect the light intensity in real time.