A method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation

Abstract

This invention relates to the field of coal pile moisture content monitoring technology, specifically a hyperspectral moisture content monitoring method for coal piles based on illumination difference compensation. The method acquires hyperspectral and elevation data from a UAV, establishes a correspondence between slope and solar incidence to extract illumination information, classifies illuminated and backlit pixels based on brightness ranking, selects stable bands to extract reflectance differences, adjusts the reflectance of various pixels according to the angle segment to generate a compensation spectrum, and matches the compensation features with the measured moisture content point by point to generate the coal pile moisture content monitoring result. This invention selects reference bands based on slope aspect and solar incidence coupling and brightness differences, constructs a multi-level illumination classification to enhance illumination recognition capabilities, selects stable bands through the correspondence between illuminated and backlit areas to construct a difference sequence, determines the adjustment direction based on the angle and position to form a consistent compensation spectrum, and matches the compensation spectrum with the moisture content to achieve full-domain mapping, improving the accuracy and spatial continuity of moisture content recognition.

Description

Technical Field

[0001] This invention relates to the field of coal pile moisture content monitoring technology, and in particular to a method for monitoring coal pile hyperspectral moisture content based on illumination difference compensation. Background Technology

[0002] The field of coal pile moisture content monitoring technology includes the detection of moisture status during coal storage and transportation, methods for measuring sample moisture content, and non-contact detection methods based on spectral information. The core of this technology lies in the quantitative characterization of the moisture content of coal and the acquisition of corresponding data through physical measurement or optical response. This includes methods such as drying measurement after manual sampling, obtaining multi-band reflectance information through hyperspectral imaging, and analyzing the spectral variation law of coal under different moisture conditions. The overall technical system covers the impact of sampling methods on representativeness, detection cycle and cost control, interference of coal pile surface spatial structure on measurement, and the reflection difference mechanism formed by the illuminated and shaded surfaces under natural lighting conditions. It also systematically studies the impact of changes in lighting conditions on the stability of spectral data.

[0003] Among them, the hyperspectral moisture content monitoring method for coal piles based on illumination difference compensation refers to a monitoring method that performs difference correction processing on hyperspectral reflectance data acquired from different illumination areas when there are illuminated and shaded surfaces on the coal pile surface. This involves setting sampling points in areas with different slopes and orientations of the coal pile, acquiring reflectance data of the illuminated and shaded areas under multiple continuous bands, recording the illumination intensity and incident angle parameters at the corresponding acquisition time, and comparing and analyzing the spectral response of the same coal body under different illumination conditions. By establishing the correspondence between the spectrum of the illuminated surface and the spectrum of the shaded surface, the reflectance of each band is corrected point by point. At the same time, combined with the absorption characteristics of moisture in specific bands, the correspondence between the corrected spectral data and the moisture content is matched to complete the monitoring process of coal pile moisture content.

[0004] Existing technologies rely on establishing spectral correspondences between overall illumination conditions and local sampling point records, lacking pixel-level geometric correlation constraints. This makes it difficult to accurately characterize illumination differences in scenarios with significant changes in slope direction, leading to aliasing in reflectance distribution. Partitioning methods are mostly based on experience or limited sample point divisions, making it difficult to form a stable spatial structure. In complex terrain, the boundaries between illuminated and transitional areas are often unclear. In continuous band processing, response stability is not distinguished, and changes in illumination can easily cause fluctuations in differences between bands, affecting the consistency of feature extraction. In the compensation processing stage, the lack of angle segmentation and spatial weight constraints leads to deviations in reflectance adjustment between shaded and weakly illuminated areas, reducing the reliability and spatial consistency of moisture content determination results. Summary of the Invention

[0005] To address the technical problems existing in the prior art, this invention provides a method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation. The technical solution is as follows: A method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation includes the following steps: S1: Acquire multi-band image data of the coal pile surface, solar altitude angle, solar azimuth angle and elevation information, determine the slope direction and solar incidence direction of each pixel, extract the brightness information of each band pixel and compare the distribution differences, select the reference band and associate the brightness, slope direction and solar incidence direction information to generate coal pile light information. S2: Based on the slope direction, solar incidence direction and brightness information in the coal pile light-receiving information, sort the brightness of pixels with the same slope direction and divide them into upper, middle and lower intervals. Combine the corresponding relationship to identify the light-receiving pixels, transition pixels and backlight pixels, remove pixels outside the coal pile boundary range, and generate a coal pile light-receiving partition sequence. S3: Based on the coal pile light-receiving partition sequence, the range of reflectance variation of the light-receiving pixels and back-light pixels in each hyperspectral band is statistically analyzed. The correspondence of the same hyperspectral band is compared, stable bands are selected, and the difference in reflectance variation of the light-receiving pixels and back-light pixels at the same coal pile location in the stable band is extracted to generate a coal pile illumination difference sequence. S4: Call the coal pile illumination difference sequence and coal pile light-receiving information, determine the included angle value and divide the angle segment, classify the lit pixels, backlit pixels and transition pixels into the corresponding angle segment, combine the spatial position relationship of various pixels and reflectance adjustment direction, determine the compensation reflectance of each pixel in the stable band, and generate the coal pile compensation spectrum record.

[0006] As a further aspect of the present invention, the coal pile illumination information includes pixel illumination intensity values, illumination direction consistency identifiers, brightness distribution characteristic values, and illumination spatial correlation attributes; the coal pile illumination partition sequence includes illuminated pixel set identifiers, transition pixel set identifiers, backlit pixel set identifiers, and partition spatial distribution structure; the coal pile illumination difference sequence includes band reflectance difference values, band response change curves, stable band identifier sets, and difference sequence continuity characteristics; and the coal pile compensation spectral record includes compensation reflectance values, angle segment division identifiers, reflectance correction parameters, and spatial transition correlation weights.

[0007] As a further aspect of the present invention, the step of obtaining S1 is as follows: S11: Acquire multi-band image data, solar altitude angle, solar azimuth angle, and coal pile surface elevation information from the digital elevation model by a drone equipped with a hyperspectral camera on the coal pile surface. Organize the slope distribution of each pixel on the coal pile surface based on the coal pile surface elevation information, determine the slope direction of each pixel according to the pixel position correspondence, establish the correspondence between pixel position and slope direction, and obtain the slope direction sequence. S12: Based on the slope direction sequence, call the solar altitude angle and solar azimuth angle to organize the solar incidence direction, and match the slope direction with the solar incidence direction item by item according to the pixel position. According to the corresponding state, distinguish between positive correspondence, lateral correspondence and negative correspondence, establish a record of the direction relationship between pixel position, and obtain the direction association sequence. S13: Based on the directional association sequence, extract the brightness information of each band pixel in the multi-band image data in sequence, organize the brightness distribution state of each band around different directional relationships, select the band with more stable directional distinction state as the reference band, and associate the brightness information of the reference band with the slope direction information and the solar incidence direction information pixel by pixel to generate the coal pile light reception information.

[0008] As a further aspect of the present invention, the step of obtaining S2 is as follows: S21: Based on the slope direction information, solar incidence direction information and brightness information of each pixel in the coal pile light-receiving information, merge the corresponding pixels according to the slope direction, arrange the brightness arrangement order along the same slope direction, divide the upper interval, middle interval and lower interval according to the arrangement position, and write the sorting interval belonging mark to each pixel to obtain the brightness layered interval. S22: Based on the brightness layering interval, call the slope direction information and solar incident direction information of each pixel, check the direction correspondence status item by item, classify the positive corresponding and upper interval pixels into the light-receiving class, classify the side corresponding and middle interval pixels into the transition class, classify the back corresponding and lower interval pixels into the backlight class, and write the light-receiving attribute mark to each pixel to obtain the light-receiving attribute mark; S23: For the light-receiving attribute markers, obtain the elevation information of the coal pile surface in the digital elevation model, organize the coal pile boundary range along the elevation distribution, check the corresponding status of each pixel position and the boundary range item by item, remove pixels outside the boundary range, retain the light-receiving, transition, and backlighting markers within the boundary range, and organize the partition records according to the pixel position order to generate the coal pile light-receiving partition sequence.

[0009] As a further aspect of the present invention, the step of obtaining S3 is as follows: S31: Based on the division results of the illuminated pixels and backlit pixels in the light-receiving partition sequence of the coal pile, retrieve the reflectance records of each hyperspectral band of the multi-band image data of the coal pile surface, collect the reflectance of the illuminated pixels and the backlit pixels according to the band, organize the start and end states of the reflectance fluctuation of each band along the pixel position, form the corresponding fluctuation interval of the two types of pixels under the same band, and obtain the reflectance fluctuation interval. S32: Based on the reflectivity fluctuation range, check the corresponding status of the fluctuation range of the light-receiving pixel and the fluctuation range of the backlight pixel in the same hyperspectral band, record the order of the beginning and end of the range and the connection status, retain the bands with consistent order and continuous connection, arrange the corresponding marks along the bands, and collect the continuously retained band numbers to obtain a stable band set. S33: Based on the set of stable bands, retrieve the reflectance of the illuminated pixels and the reflectance of the backlit pixels at the same coal pile location, organize the reflectance difference between the two types of pixels along each stable band, connect the difference records end to end according to the order of the hyperspectral bands, and write the continuously arranged content to the corresponding coal pile location to generate a coal pile illumination difference sequence.

[0010] As a further aspect of the present invention, the process of organizing the start and end states of reflectivity fluctuations in each band along the pixel position specifically includes: The reflectance of the light-receiving pixel and the reflectance of the backlight pixel are arranged according to the spatial position of the pixel. The reflectance at both ends of the arrangement sequence is extracted as the start position and end position of the fluctuation. The start position and the end position of the fluctuation are written into the recording unit of the same hyperspectral band to form a reflectance fluctuation range. The process of recording the beginning order, ending order, and connection status of the recording range is as follows: The reflectance fluctuation range of the light-receiving pixel and the reflectance fluctuation range of the backlight pixel within the same hyperspectral band are checked item by item. The order of the first end and the order of the last end are determined according to the order of the start and end positions of the two types of fluctuation ranges. The connection state is determined based on the continuous correspondence of the two types of fluctuation ranges in reflectance change.

[0011] As a further aspect of the present invention, the process of systematically organizing the reflectivity differences of the two types of pixels along each stable band specifically involves: According to the band arrangement order in the stable band set, the reflectance of the light-receiving pixel and the reflectance of the backlight pixel corresponding to the same coal pile location are called one by one, and the reflectance difference of the corresponding bands is written into the continuous recording unit in sequence. The content is then connected end to end according to the band order to form a continuous arrangement.

[0012] As a further aspect of the present invention, the step of obtaining S4 is as follows: S41: Call the reflectivity variation difference of each stable band in the coal pile illumination difference sequence and the slope direction information and solar incident direction information in the coal pile illumination information, check the corresponding state of the slope direction and solar incident direction for each pixel, organize the angle of each pixel and divide the angle segment according to the angle range, and write the corresponding angle segment assignment to the illuminated pixel, backlit pixel and transition pixel to obtain the angle segment assignment map; S42: According to the angle segment assignment map, check the position distribution of the light-receiving pixel and the backlight pixel along each angle segment, sort out the position order of the transition pixel between the light-receiving pixel and the backlight pixel, combine the difference in reflectivity change of each stable band, determine the direction of reflectivity change of the light-receiving pixel and the direction of reflectivity change of the backlight pixel respectively, and write the band adjustment mark to the corresponding position to obtain the band adjustment mark; S43: Based on the band adjustment mark, call the transition pixel position order along each angle segment, connect the reflectance change direction of the illuminated pixel and the reflectance change direction of the backlit pixel, stabilize the transition reflectance of the transition pixel one by one, and collect the corresponding reflectance of the illuminated pixel, the backlit pixel and the transition pixel to the position of each pixel band to generate a coal pile compensation spectrum record.

[0013] As a further aspect of the present invention, the method further includes: S5: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record in the stable band, collect the coal pile surface moisture content measurement data and determine the pixel position corresponding to the measurement point, extract the compensated reflectance features and compare their matching with the moisture content measurement results, screen the judgment criteria and match them point by point, and generate the coal pile moisture content monitoring results. The monitoring results of the coal pile moisture content include pixel moisture content values, spatial distribution map of moisture content, moisture content level classification, and abnormal moisture content area identification.

[0014] As a further aspect of the present invention, the step of obtaining S5 is as follows: S51: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record on the stable band, collect the measurement data of the moisture content on the coal pile surface, verify the corresponding pixel number by combining the spatial location of the measurement point, organize the compensated reflectance performance of the corresponding pixel of the measurement point along the stable band, and write the measurement results and pixel number sequence into the same recording unit to obtain the spectrum corresponding to the measurement point. S52: According to the spectrum corresponding to the measurement point, check the corresponding state of the compensated reflectance performance and the moisture content measurement result point by point, compare the reflectance fluctuation order and moisture content distribution order of each measurement point along the stable band, retain the continuous and consistent content of the corresponding state, remove the disconnected content of the corresponding state, and write the discrimination basis mark to the corresponding band position to obtain the discrimination basis mark. S53: Based on the discrimination criteria marker, retrieve the corresponding pixel compensation reflectance performance of the measurement point and match it with the moisture content measurement result item by item, organize the moisture content attribution record of each measurement point, and then check the corresponding status of the compensation reflectance performance of all pixels with the discrimination criteria marker along the stable band, write the moisture content attribution content to each pixel, and generate the coal pile moisture content monitoring result.

[0015] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following: In this invention, a unified illumination representation benchmark is formed by introducing a pixel-level coupling relationship between the slope direction and the solar incidence direction and selecting reference bands based on brightness differences. A multi-level illumination classification is constructed based on the brightness sequence of the same slope direction, and a spatial distribution structure is established to enhance the illumination recognition capability of complex surfaces. Stable bands are selected by the correspondence between the reflected and backlight reflectances, and a continuous difference expression sequence is constructed to reduce the impact of band response fluctuations. The reflectance adjustment direction and transition weight are determined by combining the angle segmentation constraint and the spatial position relationship to form a consistent compensation spectral expression. Based on the compensation spectrum and the water content data, discrimination features are extracted and global mapping is achieved, thereby improving the accuracy and spatial continuity of water content distribution recognition under different illumination conditions. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the acquisition process of S1 in this invention; Figure 3 This is a flowchart illustrating the acquisition process of S2 in this invention; Figure 4 This is a flowchart illustrating the acquisition process of S3 in this invention; Figure 5 This is a flowchart illustrating the acquisition process of S4 in this invention; Figure 6 This is a flowchart of the acquisition process for S5 of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the difference between them, their intended meanings are consistent. Similarly, the terms "of," "correlation (ponding)," and "correlation (ponding)" may sometimes be used interchangeably. It should be noted that, without emphasizing the difference between them, their intended meanings are consistent.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a technical solution: a method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation, comprising the following steps: S1: Acquire multi-band image data, solar altitude angle, solar azimuth angle, and coal pile surface elevation information from the digital elevation model, collected by a drone equipped with a hyperspectral camera on the coal pile surface. Based on the coal pile surface elevation information, determine the slope direction of each pixel on the coal pile surface. Determine the solar incidence direction based on the solar altitude angle and solar azimuth angle. Establish the correspondence between the slope direction and the solar incidence direction with each pixel as the corresponding unit. Sequentially extract the pixel brightness information of each band in the multi-band image data. Compare the distribution differences of pixel brightness information of each band under different correspondences between the slope direction and the solar incidence direction. Select a reference band, acquire the brightness information of each pixel in the reference band, and associate it with the slope direction information and the solar incidence direction information to generate coal pile light-receiving information. S2: Based on the slope direction information, solar incidence direction information, and brightness information of each pixel in the coal pile's light-receiving information, sort the brightness information of pixels with the same slope direction, and divide the upper, middle, and lower intervals according to the sorting results. Determine the sorting interval to which each pixel's brightness information belongs. Combine the correspondence between the slope direction and the solar incidence direction to complete the pixel's light-receiving attribute discrimination. For pixels whose slope direction and solar incidence direction are positively correlated and whose brightness information is located in the upper interval, they are classified as light-receiving pixels. For pixels whose slope direction and solar incidence direction are laterally correlated and whose brightness information is located in the middle interval, they are classified as transition pixels. For pixels whose slope direction and solar incidence direction are oppositely correlated and whose brightness information is located in the lower interval, they are classified as backlit pixels. Determine the coal pile boundary range according to the digital elevation model, remove pixels located outside the coal pile boundary range, and generate a coal pile light-receiving partition sequence. S3: Based on the division results of illuminated and backlit pixels in the coal pile illumination zoning sequence, for each hyperspectral band in the multi-band image data of the coal pile surface, the reflectance variation range of illuminated and backlit pixels in each hyperspectral band is statistically analyzed. The correspondence between the reflectance variation range of illuminated and backlit pixels in the same hyperspectral band is compared. Bands where the reflectance variation range of illuminated and backlit pixels remains consistent are selected as stable bands. Within the stable bands, for the illuminated and backlit pixels corresponding to the same coal pile location, the reflectance variation difference of each stable band is extracted. The bands are then continuously spliced ​​in the order of hyperspectral bands to generate a coal pile illumination difference sequence. S4: Call the reflectance variation differences of each stable band in the coal pile illumination difference sequence and the slope direction information and solar incident direction information in the coal pile illumination information to determine the angle between the slope direction and solar incident direction of each pixel. Divide the angle segments according to the range of the angle values ​​and assign the illuminated pixels, backlit pixels and transition pixels to the corresponding angle segments. Determine the distribution position of the illuminated pixels and backlit pixels in each angle segment, as well as the spatial positional relationship of the transition pixels relative to the illuminated pixels and backlit pixels. In each angle segment, based on the reflectance variation differences of each stable band, determine the reflectance adjustment direction of the illuminated pixels and backlit pixels in the stable band. Combine the spatial positional relationship of the transition pixels to determine the transition reflectance of the transition pixels in the stable band. Obtain the compensation reflectance of each pixel in the stable band and generate the coal pile compensation spectral record. S5: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record in the stable band, collect the moisture content measurement data of the coal pile surface, determine the pixel position corresponding to the moisture content measurement point, extract the compensated reflectance features of the corresponding pixel in the stable band, compare the matching of the compensated reflectance features with the moisture content measurement results, screen the compensated reflectance features as the basis for moisture content discrimination, perform point-by-point matching of the compensated reflectance features of the corresponding pixel in the moisture content measurement point with the moisture content measurement results, determine the moisture content discrimination result of the corresponding pixel in the measurement point, and determine the moisture content of each pixel in the stable band based on the point-by-point matching results, and generate the coal pile moisture content monitoring results.

[0023] The coal pile illumination information includes pixel illumination intensity values, illumination direction consistency indicators, brightness distribution characteristic values, and illumination spatial correlation attributes. The coal pile illumination zoning sequence includes illuminated pixel set identifiers, transition pixel set identifiers, backlit pixel set identifiers, and zoning spatial distribution structure. The coal pile illumination difference sequence includes band reflectance difference values, band response change curves, stable band identifier sets, and difference sequence continuity characteristics. The coal pile compensated spectral record includes compensated reflectance values, angle segment division identifiers, reflectance correction parameters, and spatial transition correlation weights. The coal pile moisture content monitoring results include pixel moisture content values, moisture content spatial distribution maps, moisture content grade division identifiers, and abnormal moisture content area identifiers.

[0024] Please see Figure 2 The steps to obtain S1 are as follows: S11: Acquire multi-band image data, solar altitude angle, solar azimuth angle, and coal pile surface elevation information from the digital elevation model by a drone equipped with a hyperspectral camera on the coal pile surface. Organize the slope distribution of each pixel on the coal pile surface based on the coal pile surface elevation information, determine the slope direction of each pixel according to the pixel position correspondence, establish the correspondence between pixel position and slope direction, and obtain the slope direction sequence. The acquisition of multi-band imagery data, solar altitude angle, solar azimuth angle, and coal pile surface elevation information from a digital elevation model (DEM) was achieved by acquiring these data from a UAV equipped with a hyperspectral camera on the coal pile surface. The basic data acquisition was performed via a low-altitude flight mission using a rotary-wing UAV equipped with a pushbroom hyperspectral imager and a lidar scanner. The multi-band imagery data was radiometrically corrected using a radiometric calibration plate to convert it into accurate surface reflectance data. The DEM was generated by interpolating lidar point cloud data. The solar altitude angle and solar azimuth angle were calculated using an astronomical ephemeris algorithm combined with GPS latitude and longitude information at the time of acquisition. The elevation values ​​of individual pixels in the coal pile DEM were extracted. Indexed by cell coordinates Establish the matrix distribution state and call the elevation difference of each pixel and its adjacent pixels; Calculate the slope variation component of this pixel in the east-west direction. and the slope variation component in the north-south direction. ,in and The ground spatial resolution of the hyperspectral image was set to 0.05 meters. The calculated dimensionless components were then substituted into the slope direction calculation formula to obtain the slope azimuth angle values ​​for each pixel. Setting true north as 0 degrees, rotating clockwise 360 ​​degrees, and calculating the actual elevation data for pixels on a certain side of the coal pile's slope, as shown below... The value is 0.5. The value is 0.866, passed. The calculated slope angle of this pixel is 60 degrees. This angle value is then compared with the corresponding pixel position coordinates. The corresponding data is stored to form a set of azimuth angles covering the entire surface of the coal pile. For flat areas or abnormally protruding areas at the edge of the coal pile, the results are verified. and If the value approaches 0, it indicates that the area does not possess effective slope angle characteristics. Therefore, the slope direction of such pixels is forcibly recorded as an invalid value or a flat land marker, and they are masked in subsequent extraction processes and not included in the light and shadow distinction. By traversing all pixel positions in the image, the row and column numbers of each coordinate point are compared with the effectively calculated slope azimuth angle. A mapping table is established to obtain a slope direction sequence containing the geographic pointing features of all pixels.

[0025] S12: Based on the slope direction sequence, call the solar altitude angle and solar azimuth angle to organize the solar incidence direction, and match the slope direction with the solar incidence direction item by item according to the pixel position. According to the corresponding state, distinguish between positive correspondence, lateral correspondence and negative correspondence, establish a record of the direction relationship between pixel position, and obtain the direction association sequence. Based on the slope direction sequence, the solar altitude angle calculated according to the astronomical calendar at the time of acquisition is retrieved. 45 degrees and solar azimuth angle The effective slope azimuth angle is obtained by setting the slope direction to 180 degrees and then using this angle to obtain the effective slope azimuth angle corresponding to each pixel position in the slope direction sequence. By performing subtraction and absolute value calculation operations on a pixel-by-pixel basis, the absolute value of the angle between the slope direction of each pixel and the solar azimuth angle is calculated. Based on the absolute value range of the included angle, a criterion for determining the direction correspondence is set. The criterion value for the positive correspondence interval is set to 0 degrees to 60 degrees, the criterion value for the lateral correspondence interval is set to 60 degrees to 120 degrees, and the criterion value for the negative correspondence interval is set to 120 degrees to 180 degrees. Substituting this into a specific calculation example, if the slope azimuth angle of a certain pixel... The angle is 170 degrees. Subtracting the solar azimuth angle of 180 degrees yields an absolute value of 10 degrees. Since 10 degrees falls strictly within the positive correspondence range of 0 to 60 degrees, this pixel is marked as a positive correspondence. If the slope azimuth angle of another pixel... If the azimuth angle is 80 degrees and the absolute value of the included angle is 100 degrees, and it falls within the lateral corresponding range of 60 degrees to 120 degrees, then it is marked as a lateral corresponding state. If the slope azimuth angle... The angle is 30 degrees. Due to the characteristics of absolute value calculation, the absolute value of the included angle is 150 degrees, which falls within the range of 120 to 180 degrees. It is marked as a reverse-facing state. The azimuth data of each pixel in the sequence is extracted one by one, and the above difference calculation is performed. It is then compared with the fixed physical interval for judgment. Each coordinate point is... The corresponding state results are recorded as discrete numerical codes, where 1 represents forward, 2 represents sideways, and 3 represents backward. For flat land pixels that have not been recorded with a valid slope direction, code 0 is filled in as an invalid item, forming a comprehensive association record of the logical relationship between pixel position and solar illumination angle, and obtaining the direction association sequence.

[0026] S13: Based on the direction association sequence, extract the brightness information of each band pixel in the multi-band image data in sequence, organize the brightness distribution state of each band around different direction relationships, select the band with more stable direction distinction state as the reference band, and associate the brightness information of the reference band with the slope direction information and the solar incidence direction information pixel by pixel to generate coal pile light information. Based on the directional correlation sequence, the pixel gray values ​​or initial reflectance in each hyperspectral band from 400 nm to 1000 nm are sequentially extracted from the calibrated coal pile multi-band image data. For each independent band, extract the brightness values ​​of all pixels belonging to code 1 (the positive state) in the direction association sequence, and calculate their mean by performing an arithmetic mean operation. Simultaneously extract the brightness values ​​of all pixels belonging to code 3 (i.e., the opposite state) and calculate their average value. The ratio between two means is obtained by performing a division operation. This is used as the core criterion for stability evaluation of direction-distinguishing states, and a stability determination coefficient is set accordingly. ,when If a value deviates most from the reference value of 1.0 in a certain band and the standard deviation of brightness fluctuation extracted between adjacent flight zones in that band is less than the preset threshold of 0.05, that band is determined to be a reference band. If all bands are found to have a standard deviation of fluctuation greater than 0.05, a fault-prevention action is performed, relaxing the standard deviation threshold to 0.10 and re-searching to ensure that valid bands can be selected. In a practical example, at the 750 nm band, the average brightness of the front-facing pixels is measured to be 200, and the average brightness of the back-facing pixels is 50. The ratio... The value is 4.0, while at the 900 nm band, the forward mean is 180 and the backward mean is 90, with a ratio of... The value is 2.0, because the brightness difference contrast is more pronounced in the 750 nm band. The value was the highest, and the standard deviation of the data distribution in this batch in this band was 0.02, clearly below the strict threshold of 0.05. Therefore, the 750 nm band was selected as the sole reference band, and the brightness values ​​of all pixels in the entire layer were extracted in this band. Compare it with the corresponding slope direction angle Solar altitude angle and solar azimuth The data is structured and combined to form a set of four-tuple information, which generates the light-receiving information of the coal pile.

[0027] Please see Figure 3 The steps to obtain S2 are as follows: S21: Based on the slope direction information, solar incidence direction information and brightness information of each pixel in the coal pile light-receiving information, merge the corresponding pixels according to the slope direction, arrange the brightness arrangement order along the same slope direction, divide the upper interval, middle interval and lower interval according to the arrangement position, and write the sorting interval belonging mark to each pixel to obtain the brightness layered interval. Based on the slope direction information, solar incidence direction information, and brightness information of each pixel in the coal pile's light-receiving information, the coordinates of each pixel are extracted. Corresponding slope azimuth value Solar altitude angle 45 degrees, solar azimuth angle The brightness values ​​are at 180 degrees and at a reference wavelength of 750 nanometers. Those with the same slope azimuth angle The pixels are logically clustered and merged, and the allowable range of azimuth angle fluctuation is set to... To accommodate small-scale terrain undulations, for the clustered set of pixels along the same slope direction, the brightness values ​​of all pixels within the set are extracted and sorted in descending order using a quicksort algorithm. The total number of valid pixels within the current slope direction set is then calculated. The permutation sequence is rigidly divided into three intervals based on the set proportional threshold parameter, with the upper interval boundary set to the order of rank 1 to 1. The middle interval boundary is set to the sorting position. to The lower interval boundary is set to the sorting position. to Substituting into a practical example, if the slope azimuth is 60 degrees ( The set of pixels (in degrees) contains 1000 pixels. The brightness values ​​of these 1000 pixels in the 750 nm band are extracted. After sorting, the brightness values ​​of the pixels ranked 1st to 300th are in the range of 180 to 240, which are classified as the upper range and marked with the numerical value "High". The brightness values ​​of the pixels ranked 301st to 700th are in the range of 110 to 179, which are classified as the middle range and marked with the numerical value "Mid". The brightness values ​​of the pixels ranked 701st to 1000th are in the range of 40 to 109, which are classified as the lower range and marked with the numerical value "Low". The set of slope azimuth angles in all directions is traversed and the above sorting and fixed quantile interval division are repeated. The corresponding interval classification results are associated and bound one-to-one with the original pixel coordinates to obtain the brightness layer intervals.

[0028] S22: Based on the brightness layering interval, call the slope direction information and solar incident direction information of each pixel, check the corresponding state of each direction, classify the positive corresponding and upper interval pixels into the light-receiving class, classify the side corresponding and middle interval pixels into the transition class, classify the back corresponding and lower interval pixels into the backlight class, and write the light-receiving attribute mark to each pixel to obtain the light-receiving attribute mark; Based on the brightness stratification intervals, the interval markers "High", "Mid", or "Low" corresponding to the coordinates of each pixel are extracted. Simultaneously, the numerical codes recorded in the direction association sequence are called, extracting the forward correspondence (code value 1), the lateral correspondence (code value 2), and the backward correspondence (code value 3). A cross-Boolean logic matching check is performed on the interval marker and direction code of each pixel. The first matching item check is executed. If the direction code of the pixel is determined to be 1 and the interval marker is determined to be "High", then the light-receiving attribute field of that coordinate point is assigned the value "Illuminated", representing that the point is on a strongly lit surface and the brightness feedback conforms to a high-value distribution. Substituting this into a specific calculation example, the coordinate point... The slope angle and the sun's azimuth are 10 degrees, which is a positive correspondence. Furthermore, the slope's brightness ranks in the top 20% in descending order, satisfying the first matching condition. Therefore, it is marked as a "light-receiving" point, and the second matching condition is checked. If the pixel's direction code is determined to be 2 and the interval label is determined to be "Mid," then the light-receiving attribute field of that coordinate point is assigned the value "Transition," indicating that the point is in the side-shielded area of ​​the sun and its brightness feedback is at a moderate level. If the coordinate point... If the angle between the points is 100 degrees and the brightness ranking is 500th (the middle region), then it is marked as a transitional type, and the third matching item verification is performed. If the pixel's direction code is determined to be 3 and the interval label is determined to be "Low", then the light-receiving attribute field of the coordinate point is assigned the value "Shadow", indicating that the point is in the area away from the sun and the brightness feedback is in a low-position distribution. If the coordinate point If the included angle is 150 degrees and the brightness ranking is within the last 30%, it is marked as a backlight. For the other abnormal pixels that do not meet the above three strong correlation logic (such as those that are facing away but have high brightness), they are uniformly classified into the atypical light-receiving class and written with the null value "Null". They are then masked in the final data output to obtain the light-receiving attribute label.

[0029] S23: For the light-receiving attribute markers, obtain the elevation information of the coal pile surface in the digital elevation model, organize the coal pile boundary range along the elevation distribution, check the corresponding status of each pixel position and the boundary range item by item, remove pixels outside the boundary range, retain the light-receiving, transition, and backlighting markers within the boundary range, and organize the partition records according to the pixel position order to generate the coal pile light-receiving partition sequence. For the light-receiving attribute markers, the absolute elevation values ​​of each effective pixel in the digital elevation model obtained by the airborne lidar scan are retrieved. By setting a threshold for the rate of change of elevation gradient The elevation data of each pixel is extracted, and the first-order difference change in elevation between the pixel and its eight neighboring pixels is calculated. A boundary judgment is then performed. When three consecutive pixels are judged to have a first-order difference below 0.15 and their absolute elevation values ​​are close to the ground reference elevation of 10.5 meters (determined through actual measured elevation at the site edge), these pixels are identified as tangent points at the bottom edge of the coal pile. By calling the convex hull algorithm or polygon closure algorithm in computational geometry, the coordinates of all identified tangent points are connected to form a closed polygon representing the actual geometric boundary of the coal pile. The set of all coordinate points inside this polygon boundary is then extracted. Search for the row and column numbers of each pixel in the light-receiving attribute marker one by one. The topological inclusion relationship comparison is performed using ray casting. If the pixel coordinates are determined to be outside the boundary polygon (such as surrounding roads or bare ground), a forced data removal action is performed, and all attribute tags at that location are cleared. Simultaneously, the data link call of the removed pixel is cut off in the system cache data column to ensure that it is completely excluded from the subsequent spectral extraction and interpolation compensation processes in S31 to S43, avoiding invalid data outside the boundary from interfering with the model accuracy. If the pixel coordinates are determined to be inside the boundary polygon or happen to be on the boundary line, and its attribute is not "Null", then its corresponding "Illuminated", "Transition", or "Shadow" attribute tags are retained. In the example, the coordinates of a pixel at the edge of the coal pile are... Its elevation value is 10.55 meters. Topologically, it is determined to be within the boundary and its "Illuminated" attribute is retained. The pixel coordinates of the surrounding hardened road surface... The elevation value was 10.51 meters, but it was determined to be outside the polygon and was completely removed. According to the row and column scanning order of the pixels in the image, all the remaining light-receiving, transition, and backlight attribute data were repackaged into a two-dimensional data matrix to generate the coal pile light-receiving partition sequence.

[0030] Please see Figure 4 The steps to obtain S3 are as follows: S31: Based on the division results of the illuminated and backlit pixels in the coal pile illumination partition sequence, retrieve the reflectance records of each hyperspectral band of the multi-band image data of the coal pile surface, collect the reflectance of the illuminated pixels and the backlit pixels according to the band, organize the start and end states of the reflectance fluctuation of each band along the pixel position, form the corresponding fluctuation interval of the two types of pixels in the same band, and obtain the reflectance fluctuation interval. The process of organizing the start and end states of reflectivity fluctuations in each band along the pixel position is as follows: The reflectance of the illuminated pixels and the reflectance of the backlit pixels are arranged according to the spatial position of the pixels. The reflectance at both ends of the arrangement sequence is extracted as the start and end positions of the fluctuation. The start and end positions of the fluctuation are written into the recording unit of the same hyperspectral band to form the reflectance fluctuation range. Based on the division of illuminated and backlit pixels in the coal pile illumination zoning sequence, the reflectance matrix corresponding to each hyperspectral band in the multi-band image data converted to surface reflectance after radiometric calibration is retrieved. For each specific hyperspectral band, the physical reflectance values ​​of all illuminated pixels in that band are retrieved and collected based on the row and column coordinates of the pixels marked "Illuminated" in the light-receiving attribute marker, forming a one-dimensional array of light-receiving reflectance. Simultaneously, the reflectance values ​​of all backlit pixels are retrieved and collected based on the coordinates marked "Shadow", forming a one-dimensional array of backlit reflectance. The array is then arranged according to the physical spatial order of the image scan (i.e., row number). From smallest to largest, column number Rearrange these two arrays (from smallest to largest), and extract the reflectance of the first pixel at the beginning of the permutation sequence as the starting position data of the light fluctuation. The reflectance of the last pixel at the end of the sequence is extracted as the data for the termination position of light fluctuation. To characterize the reflectivity fluctuations over this spatial span, and using a practical example, at the 750 nm wavelength, the reflectivity of the first pixel after spatial arrangement of the illuminated pixels is 0.25, and the reflectivity of the last pixel is 0.32. Therefore, the structured record of the fluctuation range at both ends of the illuminated area in this wavelength band is as follows: The same spatial location indexing logic is used to extract the two ends of the backlight reflectivity array, with the first and second spatial points having a reflectivity of 0.05 as the starting position of the backlight fluctuation. The reflectance of the last spatial point is 0.08, which is taken as the termination position of the backlight fluctuation. Then the backlight range of this band is recorded as It iterates through all available bands between 400 nm and 1000 nm and repeatedly performs the first and last value extraction operation. The obtained start and end values ​​are written into the hyperspectral band index array in pairs to obtain the reflectance fluctuation range.

[0031] S32: Based on the reflectance fluctuation range, check the corresponding status of the fluctuation range of the light-receiving pixel and the fluctuation range of the backlight pixel in the same hyperspectral band, record the order of the beginning and end of the range and the connection status, retain the bands with consistent order and continuous connection, arrange the corresponding marks along the band, and collect the continuously retained band numbers to obtain a stable band set. The process of recording the beginning order, ending order, and connection status of the range is as follows: The reflectance fluctuation range of the illuminated pixel and the reflectance fluctuation range of the backlight pixel in the same hyperspectral band are checked one by one. The order of the first end and the order of the last end are determined according to the order of the start and end positions of the two types of fluctuation ranges. The connection state is determined according to the continuous correspondence of the two types of fluctuation ranges in reflectance change. Based on the reflectance fluctuation range, extract the two endpoint values ​​corresponding to the illuminated pixels within the same hyperspectral band. The two values ​​corresponding to the backlight pixels Perform the head-end size order check and judgment action, compare and The size relationship, if If the condition is met, the reflectivity of the starting end of the illuminated pixel is determined to be greater than that of the backlight, and this is recorded as a consistent order at the beginning. Then, the size order at the end is checked and compared. and The size relationship, if If the condition is met, it is recorded as an end-consistent sequence, and a spatial change rate connection state check is performed. Assume the total number of illuminated pixels is... The total number of backlit pixels is ; Calculate the average step size of the light wave separately Average step size of backlight fluctuation ; Perform division to calculate the ratio of the two step sizes. An extreme value protection verification mechanism is added here. If the total number of light-receiving pixels involved in the calculation is determined... Or the total number of backlight pixels If the sample size is 10 or less, the step ratio will be distorted due to the small sample size. Therefore, the calculation method for this ratio will be suspended, and the Mann-Whitney U test will be used instead to perform a nonparametric test on the two sets of reflectance sequences. When the p-value is greater than 0.05, the two distributions are considered to be consistent, and a baseline value for the ideal coherence coefficient is set. The value is 1.0, and the allowable deviation threshold is set to 0.2. If the judgment is... If the conditions are met and the light received is greater than the backlight in both the first and last comparisons, then the reflectivity fluctuations in this band are determined to be continuous at the same frequency in both types of regions. These are retained and marked as having a consistent sequence and continuous connection. Substituting this into the calculation example, the light received range of the 750 nm band... Backlight area , head end And assume , ; Calculated , ,ratio If the deviation threshold is met, the band is recorded and retained. For abnormally disconnected bands that do not meet the above conditions, it means that the spectral response mechanism of the band in the light-receiving and backlighting areas has a physical break or is subject to severe signal-to-noise interference, and no longer has the basis for uniform compensation across the entire field. In this case, a forced removal action is performed and the bands are no longer involved in the calculation. All band number sets (such as 750, 800, 850 nm, etc.) that have been retained through continuity verification are summarized to obtain a stable band set.

[0032] S33: Based on the set of stable bands, retrieve the reflectance of the illuminated pixels and the reflectance of the backlit pixels at the same coal pile location, organize the reflectance difference of the two types of pixels along each stable band, connect the difference records from beginning to end according to the order of the hyperspectral bands, and write the continuously arranged content to the corresponding coal pile location to generate a coal pile illumination difference sequence. The process of sorting out the reflectance difference between the two types of pixels along each stable band is as follows: According to the band arrangement order in the stable band set, the reflectance of the light-receiving pixel and the reflectance of the backlight pixel corresponding to the same coal pile location are called one by one, and the reflectance difference of the corresponding band is written into the continuous recording unit in sequence. The content is then connected end to end according to the band order to form a continuous arrangement. Based on the stable band set, retrieve the same physical coordinate position from the hyperspectral image. The reference location includes the image data of the illuminated and backlit areas mapped to the surrounding neighborhood. To eliminate the differences in light reflection caused by different slope orientations, the average reflectance value of the illuminated surface area needs to be extracted. The average reflectance value of the same physical medium but in the backlight area The difference operation is performed one by one according to the wavelength numbering order retained in the stable band set; Calculate the absolute value of reflectivity deviation for the corresponding band. This quantitatively reflects the attenuation of spectral intensity by pure topographic shadows. Substituting this into specific numerical calculations, for the 750nm band at the beginning of the stable band set, if the reflectance of the illuminated side is 0.28 and the reflectance of the shaded side is 0.06, the difference is 0.22 after subtraction. Similarly, for the subsequent 800nm ​​and 850nm bands, the corresponding reflectance is subtracted, yielding differences of 0.24, 0.21, etc. All the calculated single-band difference results are stored in a one-dimensional linear array in ascending order of wavelength to form a difference set. After completing the orderly connection of the beginning and end, the position writing function is called to write the structured difference set into the specified position record node of the global coordinate layer of the coal pile, generating the coal pile illumination difference sequence.

[0033] Please see Figure 5 The steps to obtain S4 are as follows: S41: Call the reflectivity variation difference of each stable band in the coal pile illumination difference sequence and the slope direction information and solar incident direction information in the coal pile illumination information. Check the corresponding state of the slope direction and solar incident direction for each pixel, organize the angle of each pixel and divide the angle segment according to the angle range. At the same time, write the corresponding angle segment assignment to the illuminated pixel, the backlit pixel and the transition pixel to obtain the angle segment assignment map. By accessing the reflectivity variation differences in each stable band of the coal pile illumination difference sequence and the slope orientation and solar incidence information in the coal pile illumination information, the slope azimuth angle corresponding to each pixel is extracted. The solar altitude angle of 45 degrees and the solar azimuth angle of 180 degrees, obtained by astronomical algorithms, are used to calculate the angle between each pixel and the azimuth of the sun's incidence. This is done by subtraction and absolute value calculation. Based on the relative topological characteristics of the coal pile slope and illumination, fixed angular segment boundaries are defined. The entire angular region from 0 degrees to 180 degrees is uniformly discretized into six semi-open and semi-closed angular segments at 30-degree intervals. The first segment is... The second section is And so on up to the sixth section. Substituting the actual calculation examples, the coordinates of a certain position... The slope azimuth is extracted as 170 degrees. After calculation, the absolute value of the included angle is 10 degrees. After performing a numerical range judgment, it is assigned to the first segment. (Another coordinate...) The slope angle is 80 degrees, and the absolute value of the included angle is 100 degrees. It is determined to be classified into the fourth segment. Simultaneously, the generated lighting attribute tags of each pixel are retrieved. Under each angle segment level, a string combination writing action is performed, concatenating the attribute category with the segment number. If the pixel is a lighting type and is in the first segment, the internal label "Zone1-Ill" is written. If it is a backlight type, "Zone1-Sha" is written. If it is a transition type, "Zone1-Tra" is written. The above included angle calculation and range classification are repeated until all pixels in the layer are marked. The pixels with dual labels of angle range and lighting attribute are assembled according to the geospatial grid to obtain the angle segment classification map.

[0034] S42: According to the angle segment assignment map, check the position distribution of the light-receiving pixel and the backlight pixel along each angle segment, sort out the position order of the transition pixel between the light-receiving pixel and the backlight pixel, combine the difference in reflectance change of each stable band, determine the direction of reflectance change of the light-receiving pixel and the direction of reflectance change of the backlight pixel respectively, and write the band adjustment mark to the corresponding position to obtain the band adjustment mark. Based on the angle segment assignment map, the planar row and column coordinates of the transition pixels in the spatial grid are retrieved layer by layer according to the six angle segments. The KD-tree nearest neighbor search algorithm is called to check and extract the coordinate distribution of the centroids of lit and backlit pixels within the same angular segment, and the linear Euclidean distance from each transition pixel to the lit centroid is extracted. Euclidean distance from the center of gravity of the backlight The transition pixel position bias weighting coefficient is obtained by performing normalized division. Retrieve the reflectivity difference within the stable band To balance the difference between reflected light and backlight, the variation polarity and value of the two types of pixels were determined and set, and the original reflectance of the reflected light pixels was extracted. And subtract the leveling increment If the change in reflectance of the illuminated pixel is determined to be a negative adjustment (marked as -1), the original reflectance of the backlight pixel is extracted. And add leveling increment The backlight pixel change is determined to be positive adjustment (marked as 1). Substituting specific data examples, under a stable 750nm wavelength, the original reflectivity of a certain section is 0.28, and the difference... The value is 0.22, and the target amount of light-receiving adjustment is calculated as follows: The direction is downward, the original backlight reflectivity is 0.06, and the target adjustment value is... The direction is upward, and for transition pixels, it is based on their position weight. The interpolation adjustment amount is assigned sequentially, and the positive and negative polarities and incremental values ​​of the changes determined by different regions are encapsulated in the data structure. A hexadecimal identification code carrying the characteristics of the band and adjustment command is written at the corresponding pixel position to obtain the band adjustment mark.

[0035] S43: Based on the band adjustment mark, the transition pixel position order is called along each angle segment, the reflectance change direction of the illuminated pixel is connected with the reflectance change direction of the backlit pixel, the transition reflectance of the transition pixel is sorted out one by one in the stable band, and the reflectance of the illuminated pixel, the backlit pixel and the transition pixel are collected to the band position of each pixel to generate the coal pile compensation spectrum record. Based on band adjustment markers, spatial position bias weighting coefficients of transition pixels are extracted one by one along each fixed angle segment level. The reflectivity after light-receiving reference compensation is calculated by calling the light-receiving variation command: ; Calculate the reflectivity after backlight reference compensation by calling the backlight adjustment command: For each continuous stable band, a distance-weighted linear interpolation operation is performed to calculate the compensated reflectance of the transition pixels in the gradation region: If specific distance data is used for actual calculation, and the distance from a certain transition pixel to the centroid of light is extracted... It is 2 meters away from the center of gravity of the backlight. The value is 8 meters, and its weighting coefficient is calculated. The value is 0.8, and the calculated value for this band is extracted. and Since all values ​​are leveled to 0.17, the transition compensation reflectivity of this transition pixel is obtained through multiplication and addition: To ensure that the reflectivity benchmark of the entire slope is consistent in this band, eliminate the undulations caused by shadows and direct sunlight, and cyclically call all wavelengths in ascending order of stable band numbers to perform the above interpolation weighting operation, summarize the absolute values ​​of the lit pixels, backlit pixels and transition pixels after compensation into an independent spectral vector array, and remap this array back to the original image grid system to generate a set of uniformized standard data without light and shadow effects, and generate a coal pile compensated spectral record.

[0036] Please see Figure 6 The steps to obtain S5 are as follows: S51: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record on the stable band, collect the measurement data of the moisture content on the surface of the coal pile, verify the corresponding pixel number by combining the spatial location of the measurement point, organize the compensated reflectance performance of the corresponding pixel of the measurement point along the stable band, and write the measurement results and pixel number sequence into the same recording unit to obtain the spectrum corresponding to the measurement point. Obtain the compensated reflectance data set of each pixel in the compensated spectral record of the coal pile after eliminating illumination unevenness errors in the stable band, and extract the corrected reflectance values ​​of each effective pixel in continuous stable bands such as 750 nm and 800 nm. Simultaneously, the true percentage values ​​of moisture content at sampling points on the surface of the coal pile were obtained through on-site testing using the constant-temperature drying method strictly adhering to the national standard (GB / T211). During sampling, the three-dimensional spatial latitude and longitude coordinates recorded by a handheld RTK mapping instrument are extracted simultaneously. The geographic coordinate system transformation algorithm is then called to project these coordinates onto the remote sensing image pixel coordinate system. The subordinate relationship between the measurement point and the specific pixel row and column number is verified and determined. For example, the row and column coordinates corresponding to point A are determined. Point B corresponds to Point C corresponds to Extract the numerical sequence of compensated reflectance of point A in each stable band as follows: Then, it was bound to the moisture content of 8.5% obtained from the test, and the sequence at point B was extracted similarly. And it was bound to a measured moisture content of 12.2%, and the sequence at point C was extracted. The system binds a moisture content of 5.1%, iterates through all field measurement points to perform data collection and array splicing operations, and writes the latitude and longitude numbers of discrete sampling points, the spectral vectors of each stable band, and the measured moisture content into the same CSV format relational record cell table in a row-by-row manner, thus constructing a data dictionary that corresponds one-to-one between the measured end and the remote sensing end, and obtaining the corresponding spectrum of the measurement points.

[0037] S52: According to the spectrum corresponding to the measurement points, check the corresponding state of the compensated reflectance performance and the moisture content measurement results point by point. Along the stable band, compare the order of the reflectance fluctuations of each measurement point with the order of the moisture content distribution, retain the content with continuous and consistent corresponding states, eliminate the content with discontinuous corresponding states, and write the discrimination basis mark to the corresponding band position to obtain the discrimination basis mark; Extract the compensated reflectance of the stable band corresponding to each measurement point in the dictionary according to the spectrum corresponding to the measurement points scalars and the corresponding laboratory moisture content For a specific band such as 750 nm, the compensated reflectance values of measurement points A, B, and C are 0.17, 0.12, and 0.22 in sequence. Invoke the numerical comparison algorithm to perform an ascending sorting operation on them, and determine that the order of the reflectance from small to large is B < A < C. Synchronously extract the measured moisture contents corresponding to these three points, which are 12.2%, 8.5%, and 5.1% respectively, and perform a descending sorting operation based on the moisture content to determine that the order of the moisture content from large to small is B > A > C. Invoke the Pearson correlation coefficient formula to calculate the correlation degree between the two sets of discrete variables (where and are their arithmetic means respectively), perform a threshold comparison and judgment operation, and set the negative correlation stability judgment threshold to -0.85. If the value calculated in this band is less than or equal to -0.85 and the sorting order of the two sets of data shows a strictly consistent continuous correspondence with a complete reversal, then judge and retain this band, write the retention mark 1 to its band index position. Substitute the data example. If the value calculated in the 750 nm band is -0.92, it is judged to be highly negatively correlated and retained. If another stable band such as 850 nm has a weak background moisture absorption characteristic and its calculated value is -0.60, which is greater than the threshold, and its state is judged to be unconnected and discontinuous, then perform an elimination operation and remove this band from the inversion feature pool, write the abandonment mark 0 to it, and finally summarize and generate a data vector covering the available state Boolean values of each band to obtain the discrimination basis mark.

[0038] S53: Based on the discrimination basis mark, retrieve the compensated reflectance performance of the pixels corresponding to the measurement points and match it with the moisture content measurement results item by item, sort out the moisture content attribution records of each measurement point, and then check the corresponding state of the compensated reflectance performance of all pixels along the stable band with the discrimination basis mark, and write the moisture content attribution content to each pixel to generate the monitoring result of the coal pile moisture content; Based on the discrimination basis mark, retrieve the pixel compensated reflectance corresponding to the measurement points under the strong feature band such as 750 nm retained according to the mark bit with an internal Boolean value of 1 Perform parameter fitting and matching, select data from extreme boundary measurement points (such as measurement points A and B) in the measured dataset to establish linear slope inversion step coefficients, and calculate the slope coefficients. Substitute in the specific measured values, that is Based on this, a global linear conversion formula for pixel reflectance and coal pile moisture content is established: To ensure the reliability and robustness of the inversion formula in engineering applications, a leave-one-out cross-validation mechanism is introduced. This involves extracting remaining measurement points that did not directly participate in the calibration calculation (e.g., measurement point C, whose extracted compensated reflectance is 0.22 and the measured moisture content is 5.1%) and substituting them into the established conversion formula to calculate their predicted values, thus obtaining the predicted moisture content: The predicted value of 4.8% is compared with the actual measured value of 5.1%, and the root mean square error (RMSE) is calculated. If the RMSE is lower than the system's set allowable error threshold of 1.0%, the conversion formula is deemed to have a clear physical meaning, no overfitting, and sufficient accuracy, and can be formally used for full-field moisture content prediction and inversion. The compensated reflectance data of all effective pixels within the polygonal boundary defined in the light-receiving partition sequence are called to perform a full-coverage multiplication-addition conversion. Substituting this into the example, if the compensated reflectance of a pixel at an unmeasured point within the boundary is extracted... The value is 0.15. Substituting this value into the conversion formula, the predicted moisture content at this point is: The system sets criteria for low moisture (0%-8%), medium moisture (8%-15%), and high moisture (>15%) ranges, performs a classification comparison, and classifies the pixel into the medium moisture range because 9.98% meets the criteria. All predicted moisture percentage values ​​and their corresponding medium, high, and low classification labels are bound to spatial latitude and longitude coordinates, and the output is a visualized grid distribution layer data to generate coal pile moisture content monitoring results.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation, characterized in that, Includes the following steps: S1: Acquire multi-band image data of the coal pile surface, solar altitude angle, solar azimuth angle and elevation information, determine the slope direction and solar incidence direction of each pixel, extract the brightness information of each band pixel and compare the distribution differences, select the reference band and associate the brightness, slope direction and solar incidence direction information to generate coal pile light information. S2: Based on the slope direction, solar incidence direction and brightness information in the coal pile light-receiving information, sort the brightness of pixels with the same slope direction and divide them into upper, middle and lower intervals. Combine the corresponding relationship to identify the light-receiving pixels, transition pixels and backlight pixels, remove pixels outside the coal pile boundary range, and generate a coal pile light-receiving partition sequence. S3: Based on the coal pile light-receiving partition sequence, the range of reflectance variation of the light-receiving pixels and back-light pixels in each hyperspectral band is statistically analyzed. The correspondence of the same hyperspectral band is compared, stable bands are selected, and the difference in reflectance variation of the light-receiving pixels and back-light pixels at the same coal pile location in the stable band is extracted to generate a coal pile illumination difference sequence. S4: Call the coal pile illumination difference sequence and coal pile light-receiving information, determine the included angle value and divide the angle segment, classify the lit pixels, backlit pixels and transition pixels into the corresponding angle segment, combine the spatial position relationship of various pixels and reflectance adjustment direction, determine the compensation reflectance of each pixel in the stable band, and generate the coal pile compensation spectrum record.

2. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The coal pile illumination information includes pixel illumination intensity values, illumination direction consistency identifiers, brightness distribution characteristic values, and illumination spatial correlation attributes. The coal pile illumination zoning sequence includes illuminated pixel set identifiers, transition pixel set identifiers, backlit pixel set identifiers, and zoning spatial distribution structure. The coal pile illumination difference sequence includes band reflectance difference values, band response change curves, stable band identifier sets, and difference sequence continuity characteristics. The coal pile compensation spectral record includes compensation reflectance values, angle segment division identifiers, reflectance correction parameters, and spatial transition correlation weights.

3. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The steps for obtaining S1 are as follows: S11: Acquire multi-band image data, solar altitude angle, solar azimuth angle, and coal pile surface elevation information from the digital elevation model by a drone equipped with a hyperspectral camera on the coal pile surface. Organize the slope distribution of each pixel on the coal pile surface based on the coal pile surface elevation information, determine the slope direction of each pixel according to the pixel position correspondence, establish the correspondence between pixel position and slope direction, and obtain the slope direction sequence. S12: Based on the slope direction sequence, call the solar altitude angle and solar azimuth angle to organize the solar incidence direction, and match the slope direction with the solar incidence direction item by item according to the pixel position. According to the corresponding state, distinguish between positive correspondence, lateral correspondence and negative correspondence, establish a record of the direction relationship between pixel position, and obtain the direction association sequence. S13: Based on the directional association sequence, extract the brightness information of each band pixel in the multi-band image data in sequence, organize the brightness distribution state of each band around different directional relationships, select the band with more stable directional distinction state as the reference band, and associate the brightness information of the reference band with the slope direction information and the solar incidence direction information pixel by pixel to generate the coal pile light reception information.

4. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The steps for obtaining S2 are as follows: S21: Based on the slope direction information, solar incidence direction information and brightness information of each pixel in the coal pile light-receiving information, merge the corresponding pixels according to the slope direction, arrange the brightness arrangement order along the same slope direction, divide the upper interval, middle interval and lower interval according to the arrangement position, and write the sorting interval belonging mark to each pixel to obtain the brightness layered interval. S22: Based on the brightness layering interval, call the slope direction information and solar incident direction information of each pixel, check the direction correspondence status item by item, classify the positive corresponding and upper interval pixels into the light-receiving class, classify the side corresponding and middle interval pixels into the transition class, classify the back corresponding and lower interval pixels into the backlight class, and write the light-receiving attribute mark to each pixel to obtain the light-receiving attribute mark; S23: For the light-receiving attribute markers, obtain the elevation information of the coal pile surface in the digital elevation model, organize the coal pile boundary range along the elevation distribution, check the corresponding status of each pixel position and the boundary range item by item, remove pixels outside the boundary range, retain the light-receiving, transition, and backlighting markers within the boundary range, and organize the partition records according to the pixel position order to generate the coal pile light-receiving partition sequence.

5. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The steps for obtaining S3 are as follows: S31: Based on the division results of the illuminated pixels and backlit pixels in the light-receiving partition sequence of the coal pile, retrieve the reflectance records of each hyperspectral band of the multi-band image data of the coal pile surface, collect the reflectance of the illuminated pixels and the backlit pixels according to the band, organize the start and end states of the reflectance fluctuation of each band along the pixel position, form the corresponding fluctuation interval of the two types of pixels under the same band, and obtain the reflectance fluctuation interval. S32: Based on the reflectivity fluctuation range, check the corresponding status of the fluctuation range of the light-receiving pixel and the fluctuation range of the backlight pixel in the same hyperspectral band, record the order of the beginning and end of the range and the connection status, retain the bands with consistent order and continuous connection, arrange the corresponding marks along the bands, and collect the continuously retained band numbers to obtain a stable band set. S33: Based on the set of stable bands, retrieve the reflectance of the illuminated pixels and the reflectance of the backlit pixels at the same coal pile location, organize the reflectance difference between the two types of pixels along each stable band, connect the difference records end to end according to the order of the hyperspectral bands, and write the continuously arranged content to the corresponding coal pile location to generate a coal pile illumination difference sequence.

6. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 5, characterized in that: The process of organizing the start and end states of reflectivity fluctuations in each band along the pixel position is as follows: The reflectance of the light-receiving pixel and the reflectance of the backlight pixel are arranged according to the spatial position of the pixel. The reflectance at both ends of the arrangement sequence is extracted as the start position and end position of the fluctuation. The start position and the end position of the fluctuation are written into the recording unit of the same hyperspectral band to form a reflectance fluctuation range. The process of recording the beginning order, ending order, and connection status of the recording range is as follows: The reflectance fluctuation range of the light-receiving pixel and the reflectance fluctuation range of the backlight pixel within the same hyperspectral band are checked item by item. The order of the first end and the order of the last end are determined according to the order of the start and end positions of the two types of fluctuation ranges. The connection state is determined based on the continuous correspondence of the two types of fluctuation ranges in reflectance change.

7. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 5, characterized in that: The process of sorting out the reflectance difference between the two types of pixels along each stable band is as follows: According to the band arrangement order in the stable band set, the reflectance of the light-receiving pixel and the reflectance of the backlight pixel corresponding to the same coal pile location are called one by one, and the reflectance difference of the corresponding bands is written into the continuous recording unit in sequence. The content is then connected end to end according to the band order to form a continuous arrangement.

8. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The steps for obtaining S4 are as follows: S41: Call the reflectivity variation difference of each stable band in the coal pile illumination difference sequence and the slope direction information and solar incident direction information in the coal pile illumination information, check the corresponding state of the slope direction and solar incident direction for each pixel, organize the angle of each pixel and divide the angle segment according to the angle range, and write the corresponding angle segment assignment to the illuminated pixel, backlit pixel and transition pixel to obtain the angle segment assignment map; S42: According to the angle segment assignment map, check the position distribution of the light-receiving pixel and the backlight pixel along each angle segment, sort out the position order of the transition pixel between the light-receiving pixel and the backlight pixel, combine the difference in reflectivity change of each stable band, determine the direction of reflectivity change of the light-receiving pixel and the direction of reflectivity change of the backlight pixel respectively, and write the band adjustment mark to the corresponding position to obtain the band adjustment mark; S43: Based on the band adjustment mark, call the transition pixel position order along each angle segment, connect the reflectance change direction of the illuminated pixel and the reflectance change direction of the backlit pixel, stabilize the transition reflectance of the transition pixel one by one, and collect the corresponding reflectance of the illuminated pixel, the backlit pixel and the transition pixel to the position of each pixel band to generate a coal pile compensation spectrum record.

9. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 1, characterized in that: The method further includes: S5: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record in the stable band, collect the coal pile surface moisture content measurement data and determine the pixel position corresponding to the measurement point, extract the compensated reflectance features and compare their matching with the moisture content measurement results, screen the judgment criteria and match them point by point, and generate the coal pile moisture content monitoring results. The monitoring results of the coal pile moisture content include pixel moisture content values, spatial distribution map of moisture content, moisture content level classification, and abnormal moisture content area identification.

10. The method for monitoring the hyperspectral moisture content of coal piles based on illumination difference compensation according to claim 9, characterized in that: The steps for obtaining S5 are as follows: S51: Obtain the compensated reflectance of each pixel in the coal pile compensated spectral record on the stable band, collect the measurement data of the moisture content on the coal pile surface, verify the corresponding pixel number by combining the spatial location of the measurement point, organize the compensated reflectance performance of the corresponding pixel of the measurement point along the stable band, and write the measurement results and pixel number sequence into the same recording unit to obtain the spectrum corresponding to the measurement point. S52: According to the spectrum corresponding to the measurement point, check the corresponding state of the compensated reflectance performance and the moisture content measurement result point by point, compare the reflectance fluctuation order and moisture content distribution order of each measurement point along the stable band, retain the continuous and consistent content of the corresponding state, remove the disconnected content of the corresponding state, and write the discrimination basis mark to the corresponding band position to obtain the discrimination basis mark. S53: Based on the discrimination criteria marker, retrieve the corresponding pixel compensation reflectance performance of the measurement point and match it with the moisture content measurement result item by item, organize the moisture content attribution record of each measurement point, and then check the corresponding status of the compensation reflectance performance of all pixels with the discrimination criteria marker along the stable band, write the moisture content attribution content to each pixel, and generate the coal pile moisture content monitoring result.

Images