摘要
提出了一种新的红外图像识别算法.该算法根据图像预处理分割后的二值图像,对目标图像进行递归标记,避免了基于点的连通体检测中“I”像素点可能被重复扫描的现象,实现了只考虑提取目标特征,不用得到最终的标记图像本身的要求,更加有利于对目标的检测与跟踪.分析、实验结果表明算法对理想及带噪声的输入均有较好的识别率,能提高目标单帧检测概率.算法易于实现并行处理,提高了图像处理实时性,具有一定的工程实用性.
A new algorithm for the recognition of infrared image is proposed. Based on binary image made of image segmentation pretreatment, the recursion marking is assigned to the corresponding target. Pixel "1" being rescanned in detection of the connected set of points is avoided. The requirements of only considering abstracting the target characteristics without obtaining final marked image itself is realized. Moreover the algorithm is convenient for the target detection and the target tracking. The analysis and the experimental results show that the algorithm has good recognition rate for the ideal input and the striping noise input. As a result, the detection probability of the infrared target is improved in the single frame as far as possible. The algorithm not only easily realizes in parallel processing but also enormously enhances the real-time imagery processing. Meanwhile the algorithm has certain engineering practicability.
出处
《武汉理工大学学报(交通科学与工程版)》
2006年第6期946-949,共4页
Journal of Wuhan University of Technology(Transportation Science & Engineering)
基金
国防科技预研基金项目资助(批准号:05J3.7.2)