An Improved Snake Model And Its Application In Multiple Plane Targets Detection

At present, image processing technology is becoming widely used in military domain, exact image segmentation and edge detection are the important basis of achieving precise image guidance. Because of the features of satellite and aviatic images themselves, traditional image segmentation methods can't meet the demands and directly impact on the exact target recognition.This paper aimed at the characteristics of military images and the particular demands of military reconnaissance. It proposed the segmentation of multi-target image based on Canny operator and multi-resolution wavelet Snake Model for military image edge detection. The main ideas are as follows:(1) Particularly introduced the image detection and segmentation fundamental, especially the traditional Snake Model, and then adopted the system of plane detection using military satellite images;(2) Two-dimension discrete wavelet transformation and LUM filtering methods are adopted in the image preprocess for denoise. This preprocessing denoise method is based on wavelet frame to achieve denoise and filtering;(3) Proposed a new initialized, elasticity force computing method and a new external energy computing to improve the performance of the traditional Snake Model, and also brought forward a new external force which is called balloon force. At the same time Wavelet Pyramid technology is combined in the improved Snake Model. This can meliorate the movement trait of the Snake Model, enhance the accuracy of the target contour location;(4) Presented the segmentation method of multi-target image based on Canny operator, divided the multi-target into single ones using the central and the maximum edge segmentations in order to realize the edge detection of multi-target military images;(5) Applied the active contour model for the first time in satellite and aviatic images detection, take aeroplane and ground images as the detecting target, and get the expected results, which shows that the validity and realization of this paper's method.