| The technology of target detection in the night vision image has developed rapidly. Due to the nature of the night vision images about low resolution, low contrast, high noise, then the traditional target detection algorithm is not very effective in terms of night vision image. How through preprocessing algorithm to improve the efficiency of target detection and segmentation algorithm to detect the target are the two difficult problems in the research. In this dissertation, to solve these problems, by using infrared and LLL dual-channel sensor of night vision images to detect targets.This paper introduces the research of its background and significance, summary of target detection algorithm and steps, expanding from the pre-processing and target segmentation of night vision images. In terms of image preprocessing, researches the principles of Infrared imaging and LLL imaging, analyses their noise characteristics and imaging features, summarizes some common enhancement methods. Then to propose a contrast mixed algorithm of contrast enhancement. Improved the algorithm of adaptive median filtering, and then it is combined with wavelet theory for denoising. In addition, the LLL images are based on the temporal and spatial data fusion enhancement algorithms achieve the purpose of the pretreatment by the experimental verification. After pretreatment, the paper researches segmentation and focuses on the segmentation based on gray-level, analyses of the improved two-dimensional maximum entropy segmentation method. It selects the appropriate weights that can be achieved the purpose of suppressing section noise and background.The LLL images and infrared images of single-band both have their advantages and disadvantages. Some scenes are combined to get better detection. Propose a target detection of information fusion algorithm to be combined with the improved2-Dimensional Maximum Entropy Segmentation method and the feasibility of the algorithm is proved by system experiment. |
PDF Full Text Request