Cirrus cloud false alarm source detection method based on multiband characteristics

Abstract

The invention relates to the technical field of remote sensing image processing, in particular to a cirrus cloud false alarm source detection method based on multiband characteristics. The cirrus cloud false alarm source detection method comprises the following steps: inputting two cirrus cloud images of different bands and preprocessing the cirrus cloud images; solving the fractal dimension of apreprocessed image by using a covering method, and performing normalization processing to obtain a normalized fractal dimension feature map; enhancing the other preprocessed image by using a saliencymethod to obtain a saliency feature map; performing weighted fusion on the normalized fractal dimension feature map and the saliency feature map; and performing adaptive threshold segmentation on thefused image to obtain a detection result. The cirrus cloud false alarm source detection method method provided by the invention avoids the dependence of a machine learning method on a large amount ofsample data and the limitation of single-waveband image detection, can achieve the quick detection of a small sample, and remarkably improves the detection capability of a cirrus cloud false alarm source.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a cirrus cloud false alarm source detection method based on multi-band characteristics, computer equipment and a computer-readable storage medium. Background technique [0002] In recent years, remote sensing technology has been widely used in the fields of military, space-based detection and meteorological analysis. Among them, space infrared satellites are an important part of the earth observation and remote sensing system, and play an important role in early warning and missile interception. . For the space infrared satellite imaging system, there are a large number of false alarm sources with high radiation energy in its imaging band, mainly natural landscapes, such as high radiation terrain or phenomena such as snow mountains, frozen rivers, and high-altitude cirrus clouds. The images appear similar, with higher gray levels, which may cause false al...

AI Technical Summary

Problems solved by technology

In "Remote Sensing Applications" published in 2011, a method of using wavelet transform to detect and remove thin clouds was recorded. First, homomorphic filtering was performed on the low-frequency approximate components of thin clouds, and then wavelet reconstruction was used to detect thin clouds. However, this method It is easily interfered by ground low-frequency information, and the detection efficiency is not high; the "Journal of Surveying and Mapping" published in 2011 recorded the method of using wavelet SCM to extract texture features and combining ASM and entropy to distinguish cloud layers. Although it has certain effects, it is limited by the resolution. However, due to the influence of factors such as detectors and detectors, there is still a high error rate; the "Journal of Surveying and Mapping" published in 2015 recorded a method for cloud detection using texture features such as second-order moments and first-order differences of clouds, but the implementation of this method It is necessary to find out the interval of reliable cloud features in advance, the detection efficiency is low, and the detection speed is slow; methods such as neural network and support vector machine are heavily dependent on the completeness of training samples, as recorded in "Remote Sensing Technology and Application" published in 2013 The method of using AdaBoost classifier for cloud detection has reached more than 100,000 training samples. Not only is this method not suitable for situations where the amount of data is small, but different training needs to be carried out for different data. The requirements for feature selection is also high, resulting in low detection efficiency, and it is difficult to meet the real-time requirements

Images