Study On Surface Target Tracking Method Based On Infrared Image

In recent years Chinese aviation industry was going through a period of rapid development, andthe management of airport in safety became much more serious. The reliability of existing monitoringsystem would decrease in night and fog weather. Based on surface monitoring problem in poorvisibility, a series of tracking algorithms in infrared images were discussed to provide reference ofintelligent traffic in surface under all day and weather, and main study content was summarizedfollows:Imaging principle and property about infrared image were investigated. According to the inherentnoise and non-uniformity problem in infrared sequences, guided filtering was introduced as theprevious processing algorithm. The experimental results show that guided filtering algorithm can filternoise, adjust non-uniformity and preserve target edge and that of background, meanwhile it’s fast. Itproved that guided filtering is effective in previous processing.The Mean-Shift, a common algorithm in tracking field was present. The model of target wasbuilt based on color histogram. The Bhattacharyya parameter was introduced to evaluate the degree ofmatching. By repeatedly iteration, target center moved to new best position. Then, the color modelwas updated based on every sub-model’s contribution. The experimental results show that trackingdrift appeared in the complex situation such as target shape’s varying and clouded.According to tracking drift due to inaccurately expression of target, a precise tracking algorithmbased on local image matting was presented. Firstly, the target region was captured by user’s paintingso that local matting rectangle generated, furthermore, target and background’s representative colorset was collected. Secondly the scribble graph generated, and the matting was achieved. Finally,accurately target contour was acquired by edge detection of the alpha matte from matting, and thecolor model was updated. In experiment part, the objective comparison between Mean-Shift andmatting tracker prove that matting tracker is the better one in tracking accuracy, and matting trackercan keep tracking in some complex situation such as target clouded.At last, a tracking system based on the matting tracker was applied. The experimental resultsshows that the system function well and it can track the only target in real-time accurately.