Anti-interference aerial remote sensing image shadow accurate detection method

Abstract

The invention provides an improved shadow detection method based on a c3 space image, three aspects of aerial image enhancement, multilateral weighted smooth denoising and edge detection are improved, contrast enhancement is carried out on the c3 space image by adopting logarithmic transformation, and the pixel value range of a low-gray region is greatly expanded so as to improve the accuracy of extraction of a shadow pixel in a shadow region; de-noising processing is carried out on the image by adopting multilateral weighted filtering, and image edge details can be kept while image noise is removed; based on a Canny operator, a diagonal direction is increased to improve a gradient amplitude calculation method, a critical domain is determined by adopting adaptive double-critical domain selection, and finally edge detection is performed by adopting the double-critical domain; experiments prove that the shadow area can be successfully detected, interference of ground objects such as roofs and roads on detection results can be effectively weakened, and compared with an original algorithm before improvement, the shadow detection accuracy is obviously improved.

Description

technical field [0001] The present application relates to an accurate detection of shadows in aerial remote sensing images, in particular to an anti-jamming method for accurate detection of shadows in aerial remote sensing images, which belongs to the technical field of shadow detection in remote sensing images. Background technique [0002] Aerial remote sensing images refer to the images of ground scenes captured by aerial cameras on airplanes or other aviation vehicles. In recent years, with the continuous development of sensor technology, high-resolution aerial remote sensing images are easier to obtain and more widely used. Aerial remote sensing images can be used to study and analyze relatively microscopic spatial structures, which have the characteristics of rich spectral information, high spatial resolution, and strong flexibility. Aerial remote sensing images have a large amount of urban environmental information, and are widely used in urban economic and social dev...

AI Technical Summary

Problems solved by technology

[0008] First, when performing key feature recognition and automatic detection on aerial remote sensing images, due to the limitations of aerial imaging technology and the existence of factors such as tall and tall objects, shadows will inevitably appear in the imaging process of high-resolution aerial remote sensing images , shadows will weaken the optical and physical information of objects in the occluded area, cause interference or loss of ground object information, seriously affect the continuity, accuracy, and integrity of image data information, and give subsequent remote sensing image processing, including pattern recognition, Image matching and feature extraction bring many difficulties. Shadows will also destroy the continuity and integrity of edge information of features and affect the visualization quality and aesthetics of aerial remote sensing products. Accurate processing of shadows on remote sensing images is very important. Necessary, and how to accurately detect shadow areas is an extremely important step in shadow processing. The existing technology lacks an accurate shadow detection method for aerial remote sensing images;

[0009] Second, existing technologies based on e 3 There are some deficiencies in the space shadow detection method: one is to detect the low-saturation ground objects in the non-shaded area as shadows, and the types of ground objects in aerial remote sensing images are complex and diverse, while using c 3 When space detects shadows, the saturation and hue of asphalt roads, water bodies, vegetation and other ground objects are low, which are the same as the shadow area characteristics, in c 3 Instability in the space is often detected as shadows; the second is to detect the extremely bright or extremely small ground objects in the non-shaded area as shadows, and the extremely bright or extremely small ground objects in the aerial remote sensing image are converted to c 3 space, it has the same characteristics as the shadow area and is easily confused with the shadow area, and is often detected as a shadow; the third is to detect the blue features in the non-shadow area as shadows, and the blue features and the shadow area are in the c 3 The characteristics of the space are the same, so it is detected as a shadow area; the fourth is c 3 Spatial images are very sensitive to noise, causing pixels in shadowed areas to be detected as non-shaded areas; five is c 3 The shadow edge of the white area of ​​the space aerial remote sensing image is not obvious, and the pixels at the edge are easy to detect errors, resulting in a low detection accuracy of the shadow edge; 3 The gray scale of the spatial image is normalized to [0,255]. Although it can achieve the effect of enhancing the image contrast, the uniform transformation will lead to poor preprocessing effect;

[0010] Third, compared with other classic edge detection methods, the Canny algorithm is robust, but because aerial images may be disturbed by various factors during the process of acquisition, transmission, and conversion, the traditional Canny algorithm is used for edge detection. There will be model defects: First, the traditional Canny algorithm uses a Gaussian function to smooth and denoise the image. Since the spatial scale parameters of the Gaussian filter function are determined artificially, different sizes will have different effects on the later results, and the artificially determined parameter values ​​are very large. It is difficult to strike a balance between the signal-to-noise ratio and the precise positioning of the edge; the second is that the traditional Canny algorithm calculates the gradient amplitude by using the first-order partial derivative difference in the 2×2 neighborhood, and the calculated partial derivative does not coincide with the center of the gradient amplitude. There is an offset itself, which will have a great impact on the precise positioning of the edge; third, the high and low thresholds in the traditional Canny algorithm are artificially set, which is too subjective, and the edge detection results brought by different thresholds are also different. , in the face of different image environments, different parameter values ​​need to be adjusted, and the adaptive ability is poor;

[0011] Fourth, in the prior art, the Gaussian filter that smoothes the enhanced image to eliminate noise does not consider the edge of the image. The blurred result obtained after processing makes the overall blur of the aerial image, and the edge of the image is also blurred. The Gaussian kernel only considers The spatial distribution of pixels does not take into account the difference between pixel gray values. However, the position where the gray value of aerial images changes sharply is often the edge of the image. After processing, the general segmentation of the original image cannot be maintained and the image edge information cannot be maintained; sometimes The shadow area cannot be successfully detected, and it is very sensitive to the interference of roofs and roads on the detection results. The shadow detection anti-interference is weak, which greatly reduces the application value of shadow detection in aerial remote sensing images.

Images