Research On Technology Of Color Transferring And Target Recognition Of Infrard Video Image

With the rapid development of computer vision technology and infrared imagingtechnology, as well as the improvement of the people’s material life level. The night visionsystem based on infrared imaging which has the characteristics of intuition and convenienceand abundant content etc is increasingly brought to insteresting of study. The traditionalinfrared equipment outputs single hue video image, it is bad for the human eye observes andrecognizes target. Aiming at this problem, this paper studies the colorized processingtechnology of infrared video image. The paper turns low frequency band of gray image intohigh frequency band of color image. At the same time, due to the infrared image is highsensitivity for detecting low resolution target. By target segmentation and feature extraction tothe infrared image, the system classifies targets and aids to the human eye recognition andjudgment.Based on our previous knowledge of image processing and pattern recognition, thispaper main researches three parts which are infrared image preprocessing, color transferringof infrared image and target segmentation and target recognition of infrared image. First,aiming at infrared image’s characteristic of high background and low contrast and lowsignal-to-noise ratio. This paper mainly studies the preprocessing of infrared images anddetail enhancement technique. It analyzes and compares the image enhancement method ofhistogram equalization and histogram regulations. And it puts forward a detail enhancementmethod based on DDE technique. And then it researches a new image detail enhancementmethod based on Sobel edge detection method of eight directions. Meanwhile it improvesimage median filter denoising algorithm. Secondly, this paper’s mainly research is focused onthe color transferring algorithm of infrared image. On the basis of Welsh color transferringalgorithm, it puts forward a color transferring method based on image subsection. And itimproves the method based on color palette, the operation efficiency of the improvementmethod is enhanced. Meanwhile the idea of pseudo color and multiresolution fusion is appliedto the process of image colorization, the method can obtain colorful color infrared image.Thirdly, according to the characteristics of the target in infrared image, this paper usesmulti-threshold and self-adaptive Otsu segmentation method for segmenting the target area.And it presents an improved two-dimensional Otsu Method. Meanwhile it researches theextraction method of shape feature and motion feature, and keeping the condition of dynamicand texture and geometric features, it discusses and analyzes the color feature of target. Thenthis paper studies the target’s classification and recognition of infrared image which respectively uses template matching classifier and Bayes classifier and BP neural networkclassifier. After that it has carried on the comparative analysis in recognition rate and trainingtime and recognition time.The experiment shows that the detail enhancement method based on DDE technique andthe new image edge detail enhancement method based on Sobel detection method of eightdirections has good enhancement effect. The color image which is obtained by the imagecolor transferring and pseudocolor processing based on fusion method is availed to the humaneyes observation, in addition it can meet the needs of real-time processing. The variety offeatures are used to target recognition classifier, the recognition rate is greatly increased. Theclassifier based on BP neural network can meet require of system real-time processing inrecognition rate and recognition time. At the same time, this research used the above resultshave realized the real-time and high accuracy of classification recognition system in the formof a software package.