Research Of Remote Sensing Change Detection Based On PCNN For SAR Image

SAR is based on the principle of coherent imaging mechanism. After the ground to the scattering signal back into a remote sensing image, and it has the characteristics of long distance, through the mist of vegetation, not affected by weather conditions, all-weather, all-time Earth observation, make obtaining to the same area at different times image possible. Remote sensing image change detection is conducted on the basis of this. It aims to the same area for quantitative analysis of remote sensing image, get interested feature change information. Change detection has become one of the hot research problems in the field of the current space information science, now it has been widely used in the field of dozens. In the aspect of hydrology, drought and flood disaster monitoring, prediction and evaluation, analysis of dynamic in a glacier, sea ice movement monitoring and so on. In terms of agriculture and forestry, carries out on the grass degradation, deforestation, soil desertification dynamic monitoring, in terms of environment, environmental monitoring, and evaluation and prevention of natural disasters, etc.This paper focuses on SAR image change detection and its related technologies were discussed. The completion of the work is mainly from the following several aspects:(1)In view of the FCM algorithm is easily affected by noise and easy to fall into local optimal value, the FLICM algorithm is used for SAR change detection in this article. Ratio method is the most commonly used method to obtain difference image. The difference image is getting by neighborhood ratio method, to effectively suppress speckle noise. But there is still a lot of noise in the difference image, Gabor dictionary, based on the theory of the sparse expression be used in image de-noising problems. Finally, using FLICM algorithm to cluster denoised difference image, to obtain changed area.(2)Because relative to other model, the PCNN model has two specific characteristics of neuron, linear additive and nonlinear multiplication modulation coupling, and it has more processing power to deal with the neighborhood excitation signal and can obtain more complete change detection results of image information. To solve complex PCNN model and too much parameter, a simplified PCNN model is proposed in this paper, the parameters are reduced to two, and parameter value is obtained by experimental analysis. Through the experiments, the presented algorithm is good at noise suppression is proved, and keeps the edge information well, also has better detect ability and high detection precision.