A method for automatically extracting diffuse reflection whiteboard position and calculating reflectivity based on camera data on board

Abstract

The application belongs to the technical field of camera imaging, and discloses a diffuse reflection whiteboard position automatic extraction and reflectivity calculation method based on airborne camera data, which comprises data reading, non-uniform correction, selection of green and near-infrared wave bands to construct normalized difference index (ɑNDI) and difference value index (ɑDVI), generation of a binary mask through threshold segmentation, morphological opening and closing operation using a rectangular structural element to optimize the mask, labeling of connected regions and extraction of geometric properties including position, area and perimeter, screening of a target region according to prior morphological characteristics of the diffuse reflection whiteboard, a preset area range and an aspect ratio, positioning of a target center point and extraction of a local radiation mean value, and then calculation of reflectivity. The application realizes automatic and high-precision extraction of the diffuse reflection whiteboard position, overcomes the deficiencies of low efficiency and poor consistency of traditional manual calibration, and significantly improves the precision and operation reliability of reflectivity inversion.

Description

Technical Field

[0001] This application belongs to the field of camera imaging technology, specifically relating to an automatic extraction method for the position of a diffuse whiteboard based on airborne camera data and a method for calculating reflectance. Based on the extraction results, the method achieves accurate reflectance calculation and is suitable for the rapid production of UAV hyperspectral remote sensing reflectance products. Background Technology

[0002] With the integrated development of UAV technology and hyperspectral imaging technology, pushbroom hyperspectral cameras have become core data acquisition equipment in fields such as agricultural remote sensing, environmental monitoring, and resource exploration. When airborne hyperspectral cameras perform flight imaging missions, reflectivity coefficients need to be constructed using standard diffuse reflectance whiteboards laid on the ground in order to convert the original DN values ​​into reflectivity products. This is the core guarantee for the accuracy of reflectivity products.

[0003] However, existing technologies for locating and calculating the reflectivity of diffuse whiteboards have the following key drawbacks: 1. Manual visual interpretation is extremely inefficient: Traditional methods rely on manual searching and selecting the region of interest (ROI) of the diffuse whiteboard frame by frame in the image. In long strip imaging tasks, this is inefficient and not conducive to product automation. 2. GPS+POS marking introduces additional errors: Some solutions use ground GPS to mark the latitude and longitude of the center point of the whiteboard, and then combine it with the camera POS (Position and Attitude System) data to map it to the image. However, the superposition of GPS positioning error and POS attitude error can easily lead to the whiteboard position matching deviation, which in turn affects the accuracy of the calibration coefficient. 3. Existing automatic identification methods have a high false detection rate: Mainstream automatic identification technologies (such as spectral angle mapping and matched filtering) rely solely on spectral features, which can easily misidentify ground features with spectra similar to the whiteboard (such as light-colored building roofs, metal car roofs, and cement roads) as whiteboards, thus failing to meet the needs of engineering applications to the greatest extent. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this invention proposes an integrated processing method combining "spectral index + morphological optimization + geometric feature screening" to achieve fully automatic and high-precision extraction of the position of diffuse reflectance whiteboards, while ensuring the reliability of reflectance calculation, thus solving the problems of low efficiency, large errors, and weak anti-interference ability of traditional technologies.

[0005] This invention proposes an automated, interference-resistant method for extracting the location of a diffuse reflectance whiteboard from airborne linear array camera data. It also constructs an accurate reflectance conversion model to meet the engineering production requirements of hyperspectral reflectance products. The details of this invention are as follows: Data validity verification and reading: Read the header file (hdr) of the hyperspectral image data, obtain the core image parameters (number of bands, number of samples, total number of lines), and verify the integrity of data storage (e.g., no missing frames, no abnormal DN values); the header file data storage format must be BSQ (Band Sequential) to facilitate independent reading of target band data by band.

[0006] Non-uniformity correction: Input the camera's calibrated dark level data (Dark, i.e., the dark current grayscale value under no illumination), relative radiometric coefficients a and b, and perform radiometric correction on the effective data in step 1 to eliminate the influence of non-uniformity in the camera detector response. The correction formula is: Cal = a * (Data - Dark) + b In the formula, Cal is the gray value of the corrected image, and Data is the gray value of the original image.

[0007] Spectral index construction: Utilizing the significant spectral characteristics of the diffuse reflectance white board—high response in the green band and low response in the near-infrared band—target bands are selected from the corrected image in step 2. Green band: such as channel 40 (band40, center wavelength approximately 469.56nm); Near-infrared band: such as channel 190 (band 190, center wavelength approximately 751.99nm); Based on the two bands mentioned above, two feature indices are constructed to enhance the difference between the whiteboard and background features: Normalized Difference Index: αNDI = (band40 - band190) / (band40 + band190); Difference Index: ɑDVI = band40 - band190.

[0008] Binarization mask generation: Set a joint threshold for the αNDI and αDVI obtained in step 3. Pixels that meet the threshold condition are assigned a value of 1 (foreground, i.e., the suspected white area), and pixels that do not meet the condition are assigned a value of 0 (background), thus generating a binary mask.

[0009] Morphological optimization: The binary mask from step 4 is subjected to a combination of "opening operation → closing operation" using rectangular structuring elements. Opening operation (erosion followed by dilation): Removes isolated noise points (such as single bright pixels) smaller than the structuring element in the mask, and smooths the boundaries of the white area; Closing operation (dilation followed by erosion): fills small holes (such as dark spots caused by shadows) in the white area and connects broken local areas; The size of the rectangular structural element is "1 / 3 of the diffuse whiteboard imaging size". The conversion logic is as follows: calculate the whiteboard imaging pixel size based on the flight altitude and spatial resolution, and then take 1 / 3 (e.g., at a flight altitude of 50m and a spatial resolution of 2.5cm / pixel, a whiteboard of 50cm*50cm corresponds to 20*20 pixels, and the structural element is 7*7; at a flight altitude of 100m and a spatial resolution of 5cm / pixel, the whiteboard corresponds to 10*10 pixels, and the structural element is 3*3).

[0010] Connected region labeling: Connected region analysis is performed on the binary mask optimized in step 5. The "8-neighborhood connectivity" rule is adopted (that is, a cell connected to the cells above, below, left, right and diagonal are considered as the same region). A unique label is assigned to each connected region, and the total number of regions (NumRegions) is counted.

[0011] Geometric attribute extraction: Calculate the core geometric attributes of each connected region, including: Location information: coordinates of the minimum bounding rectangle (top left corner Row_min / Col_min, bottom right corner Row_max / Col_max), centroid coordinates (Row_centroid / Col_centroid); Morphological parameters: Area (unit: number of pixels), Perimeter (unit: cell side length).

[0012] Whiteboard area filtering: Based on the preset shape characteristics of the diffuse whiteboard, the connected regions in step 7 are filtered to retain regions that meet all of the following conditions: Area requirements: The area must be within the range of "(theoretical whiteboard size ± 2 pixels)²" (e.g., at a flight altitude of 50m and a spatial resolution of 2.5cm / pixel, a 50cm*50cm whiteboard corresponds to 20*20 pixels, i.e., the selection range is 324-484 pixels). Aspect ratio requirement: The width / height of the minimum bounding rectangle should be between 0.8 and 1.2 (i.e., close to a square, which matches the actual shape of the whiteboard). Local radiation mean extraction: For the whiteboard area determined in step 8, take a circular window with a specified radius R as the center (R = 1 / 5 of the whiteboard imaging radius, such as the whiteboard imaging size of about 20 pixels at a height of 50m, R is 4 pixels), and calculate the gray mean (meanDNi) of the Cal image within the window. Reflectance conversion coefficient construction: Using the standard reflectance data of the known diffuse white board, and combining it with the meanDNi from step 9, construct the reflectance conversion coefficient for each band: Ki = Refi / meanDNi In the formula, Ki is the conversion coefficient of the i-th band, Refi is the standard reflectance of the whiteboard in the i-th band, and meanDNi is the mean local radiation of the whiteboard in the i-th band.

[0013] Reflectance image generation: Multiply the conversion coefficient Ki from step 10 by the corrected image Cal from step 2 band by band to obtain the final reflectance image. Refdata = Ki * Cal In the formula, Refdata is a reflectance image, in dimensionless form (usually represented by 0-1). Compared with existing technologies, the advantages of the present invention's method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data are as follows:

[0014] 1. Significantly improved efficiency: The fully automated processing of diffuse whiteboard extraction and reflectivity calculation is achieved. In images with a wavelength of 200 and a resolution of 1284*6097 pixels, the processing time is 2.15 seconds, which is significantly more efficient than manual processing. 2. High positioning accuracy: Combining spectral index and geometric feature filtering, it effectively eliminates interference from similar ground features, achieving 100% positioning accuracy on existing datasets. 3. Strong engineering applicability: The algorithm has strong adaptability and consumes little computing resources, and can be integrated into UAV hyperspectral data processing systems. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data, according to the present invention. Figure 2 This is a diagram illustrating the implementation effect of the method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data, based on an airborne camera data, at a flight altitude of 50m. Figure 3 This is a diagram illustrating the implementation effect of the method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data, at a flight altitude of 100m. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0017] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] like Figure 1 The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data includes the following steps: Step 1: Read the Data header file information of the outdoor scene data to obtain basic image information, including the number of bands, the number of samples, and the total number of lines. Read the image and determine whether the data is valid. Samples = 1284, Bands = 200.

[0019] Step 2: Input dark level data Dark, relative emissivity a, and relative emissivity b, and perform non-uniform correction on the effective data from Step 1; Cal = a * (DN-Dark) + b Step 3: Based on the significant spectral characteristics of the diffuse whiteboard in a specific band, select the green band and near-infrared band (e.g., 40-channel band) for the relative radiometrically corrected image obtained in Step 2. 40 and 190-channel band 190 Construct the normalized difference index aNDI and the difference index aDVI; ɑNDI=( band 40 - band 190 ) / (band 40 + band 190 ) ɑDVI = band 40 - band 190 Step 4: Set conditional thresholds for the αNDI and αDVI obtained in Step 3 to create a binarized mask for extracting the whiteboard area. Pixels that meet the threshold conditions are assigned a value of 1 (foreground), and pixels that do not meet the conditions are assigned a value of 0 (background).

[0020] Step 5: Apply a rectangular structuring element to the binary image obtained in Step 4 for morphological operations, performing opening and closing operations. The opening operation removes isolated noise points and small objects smaller than the structuring element in the binary mask, smoothing the target boundaries. The closing operation fills in small holes and gaps within the target area of ​​the binary mask and connects adjacent areas. The structuring element size is set to 1 / 3 of the expected size. A 7x7 matrix is ​​used at a height of 50m, and a 3x3 matrix is ​​used at a height of 100m.

[0021] Step 6: Perform connected region labeling analysis on the morphologically processed binary image obtained in Step 5 to obtain the unique label of each connected region and the number of regions, NumRegions.

[0022] Step 7: Calculate the connected region attributes obtained in Step 6, including at least the region's location information (such as the coordinates of the minimum bounding rectangle and the coordinates of the centroid), the region's area (Area), and the region's perimeter (Perimeter).

[0023] Step 8: For each connected region obtained in Step 7, filter out regions that are close to the blank board shape by using geometric features. The judgment conditions include the theoretical area of ​​the region and the aspect ratio.

[0024] Among them, the area condition is: the area should be within the expected diffuse whiteboard pixel area range, and the specific speed-to-altitude ratio resolution is set according to the flight altitude and payload design. Aspect ratio: The ratio of the width to the height of the minimum bounding matrix of the region should be close to 1.

[0025] The data was obtained from a 50*50cm diffuse whiteboard at flight altitudes of 50m and 100m, corresponding to spatial resolutions of 2.5cm / pixel and 5cm / pixel, respectively. The algorithm's masking effect and interpretation results are compared with the corresponding RGB original images, as shown below. Figure 2 , Figure 3 As shown.

[0026] Step 9: Based on the whiteboard area determined in Step 8, obtain the row and column numbers (Row_center and Col_center) of the whiteboard center point. Using the center point as the core, take the mean gray value (meanDN) of a circular window area with a specified radius R on the corrected image Cal. i .

[0027] Step 10: Construct the conversion relationship K based on the known standard reflectance Ref of the diffuse whiteboard and the mean DN obtained in Step 9. i =Ref i / meanDN i .

[0028] Step 11, K calculated in step 10 iMultiplying the data by the corrected image Cal yields the reflectance image, Refdata=K i *Cal.

[0029] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data, characterized in that, Includes the following steps: Step 1: Read the header file information of the hyperspectral image data, obtain basic image information, read the image and determine whether the data is valid; the header file data storage format is BSQ. Step 2: Input dark level data, relative radiometric coefficients a and b, and perform non-uniform correction on the effective data in Step 1. The correction formula is Cal = a * (Data-Dark) + b, where Cal is the corrected image and Data is the original image. Step 3: Select the green band λ for the corrected image obtained in Step 2. green With near-infrared band λ nir Normalized difference index aNDI and difference index aDVI are constructed, wherein aNDI = (green band - near-infrared band) / (green band + near-infrared band) and aDVI = green band - near-infrared band. Step 4: Set a joint threshold for αNDI and αDVI from Step 3 to generate a binarized mask; Step 5: Perform morphological opening and closing operations on the binary mask from Step 4 using a rectangular structuring element. The size of the rectangular structuring element is 1 / 3 of the size of the diffuse whiteboard image. Step 6: Mark the connected regions of the binary mask after preprocessing in Step 5, and obtain the unique label of each region and the total number of regions; Step 7: Calculate the geometric properties of each connected region from Step 6, including location information, area, and perimeter; Step 8: For the connected components from Step 7, filter and confirm the diffuse whiteboard area using geometric features; Step 9: For the diffuse whiteboard area determined in Step 8, take a circular window with a specified radius centered on the centroid and calculate the average gray value of the corrected image within the window. Step 10: Based on the standard reflectance Ref of the diffuse whiteboard and the mean DN of the grayscale obtained in Step 9... i Constructing the reflectivity conversion coefficient K i =Ref i / meanDN i ; Step 11, based on K from step 10 i Multiplying the corrected image Cal from step 2 band by band yields the reflectance image Refdata=K. i *Cal.

2. The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data according to claim 1, characterized in that, The basic image information in step 1 includes the number of bands, the number of samples, and the total number of lines. 。 3. The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data according to claim 1, characterized in that, In step 4, pixels that meet the joint threshold are assigned a value of 1, while those that do not are assigned a value of 0.

4. The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data according to claim 1, characterized in that, Step 5, the morphological opening and closing operation, involves first performing an opening operation to remove noise, and then performing a closing operation to fill in the holes.

5. The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data according to claim 1, characterized in that, The location information in step 7 includes the coordinates of the minimum bounding rectangle and the centroid coordinates.

6. The method for automatically extracting the position and calculating the reflectivity of a diffuse whiteboard based on airborne camera data according to claim 1, characterized in that, The geometric features in step 8 include the theoretical area condition of the region and the aspect ratio.

Images