Vegetation shadow detection and compensation method for hyperspectral remote sensing image

Abstract

The invention relates to the technical field of image processing, and provides a vegetation shadow detection and compensation method for a hyperspectral remote sensing image. Comprising the following steps: 1, data acquisition: acquiring hyperspectral image data; step 2, preprocessing: preprocessing the hyperspectral image data to obtain continuous low-resolution, medium-resolution and high-resolution images; step 3, shadow detection: obtaining position distribution of a shadow area and a non-shadow area according to a detection result; 4, performing shadow compensation to obtain a shadow compensation result image; according to the method, the vegetation shadow and the shadow detection and shadow compensation method for the vegetation are researched by using the hyperspectral data from the perspective of the spatial scale effect, the influence of the spatial scale effect on the shadow and the shadow processing technology is explored, and a solution thought is provided for the adaptability of the spatial scale effect of shadow compensation.

Claims

1. A method for vegetation shadow detection and compensation in hyperspectral remote sensing images. It is characterized in that The following steps are involved: Step 1: data acquisition, obtaining hyperspectral image data; Step 2: preprocessing: preprocessing the hyperspectral image data to obtain continuous low, medium and high resolution images; Step 3: shadow detection, obtaining the position distribution of shadow areas and non-shadow areas according to the detection results; Step 4: shadow compensation to obtain a shadow compensation result image.

2. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 1, It is characterized in that The preprocessing obtains continuous low, medium and high resolution images through resampling.

3. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 2, It is characterized in that The preprocessing resamples the existing data by using the nearest neighbor method.

4. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 3, It is characterized in that The preprocessing also includes spectral band removal, radiation correction, scale conversion and spectral smoothing.

5. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 4, It is characterized in that The shadow detection includes: a. Construct a grayscale image to increase the contrast between shadow and non-shadow areas; b. Threshold segmentation to determine the shadow area and non-shadow area; c. Binarization processing to obtain binary shadow detection results.

6. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 5, It is characterized in that The constructing of the grayscale image comprises: Gray is the grayscale image obtained, R, G, B are the red band, green band, and blue band in the original image, respectively, and k is the weight for adjusting the green band: Gray=|BG|+|RG|+k×G.

7. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 6, It is characterized in that The threshold segmentation uses the Otsu threshold method to obtain the threshold, including: First, a gray value T is selected from the grayscale image as the threshold, and all pixels in the grayscale image are divided into two categories: shadow pixels and non-shadow pixels, which are represented as A and B respectively; assuming that the constructed grayscale image has L grayscale levels, the grayscale range of the non-shadow area can be represented as [1, ..., T], and the grayscale range of the shadow area can be represented as [T+1, ..., L], pi is the probability distribution of the grayscale image, and pi≥0, pi=1, N is the total number of grayscale image pixels, N={n1+n2+…+nL}, the probability of the non-shadow area and the shadow area appearing is: μ t is the average grayscale level of the grayscale image, and the average grayscale levels of A and B are: The intra-class variances of A and B are: The between-class variance is: Make M reach the maximum value, then T is the optimal threshold.

8. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 7, It is characterized in that The shadow compensation includes: The light source in the remote sensing image is divided into two parts: direct light from the sun and ambient light. The difference between the lighting conditions in the shadow area and the non-shadow area is whether there is a direct light component. The remote sensing illumination model is obtained: I=ρ(kφ d +φ e ); Where: I is the pixel gray value of the remote sensing image, φ d With φ e They are the direct light component of the pixel and the ambient light component of the pixel respectively.

9. The method for detecting and compensating vegetation shadows in hyperspectral remote sensing images according to claim 8, It is characterized in that The shadow compensation further includes: The direct light component is estimated and the estimated direct light component is compensated to the shadow pixel to restore the remote sensing information; let k be the coefficient, k is 0 in the shadow area, k is 1 in the non-shadow area, ω is the ratio of direct sunlight to ambient sunlight, and the compensation model can be constructed by estimating the pixel grayscale values ​​of different environments in the remote sensing image: Let I nshw with I shw The gray value of the non-shadow area and the darker area at the edge of the non-shadow area, the compensation coefficient ω is:

10. A vegetation shadow detection and compensation method for hyperspectral remote sensing images according to any one of claims 1 to 9, It is characterized in that It also includes the construction of a spatial scale adaptive compensation model, which specifically includes the following steps: a. Extraction of end members; b. Analysis of endmember spectral characteristics; c. Adaptive compensation coefficient construction; Let I' nshw with I' shw Respectively represent the terminal pixel grayscale values ​​of the illuminated area and the natural light area, and the compensation coefficient is:

Images