Ship Target Recognition Technology Based On Infrared Imaging Study

Based on the research of the characteristics and the processing technology of infrared images and in the background of infrared imaging monitoring/tracking technology, this article achieved the infrared imaging system to track and identify ships, researched the algorithm of ship infrared image pre-process, ship waterline detection, trace point parametric estimation, target recognition.From infrared camera to computer, the analog video must be conversed to digital video first. USB video capture board stores the infrared video signal as AVI format files which are stored in the computer, and then read each frame from the AVI file, and then do subsequent processing.Because the environment of the infrared image is complex, so the acquired image contains a lot of noise and interference. Image pre-process can reduce noise, so it is great benefit to the subsequent target recognition and tracking. The paper proposed an improving Otsu image sequence segmentation algorithm, the algorithm compress calculation by compressing gray-scale range and increasing the sample interval in the image plane. At last, the simulation results of 3 different kinds of ship infrared images show that the calculation of improved Otsu is only the 1% of the original Otsu algorithm.By analyzing the relationship and feature of the background area, the target area, the plaque area and the empty area, designed efficient plaque elimination algorithm and the empty region filling algorithm which based on mathematical morphology. The simulation results show that the algorithm is feasible and effective.About target recognition, adopted edge detection and improved Hough algorithm to detecting waterline, used skeleton extraction and genetic algorithm to extracting trace points and parameter estimation. The combined moment invariants are used for target classification and identification. Used different ship images to plait simulation software, simulation experiments and analysis proved the feasibility and effectiveness of the target recognition algorithm.