一种基于图像灰度的红外目标识别算法  

Infrared Target Recognition Algorithm Based on Image Grayscale

作　　者：高晶[1] 吴昆[1] 吴育峰[2] 陈仲华[2] 

机构地区：[1]第二炮兵工程大学 [2]96630部队

出　　处：《计算机科学》2013年第9期312-316,共5页Computer Science

基　　金：国家自然科学基金项目(61003148)资助

摘　　要：针对红外图像背景复杂、识别率不高的问题,提出了一种新的基于图像灰度的红外目标识别方法。首先在形态学理论的基础上设计了一种多级形态学目标检测算子,并结合区域生长法检测目标区域;其次在候选目标区域的基础上引入sigmoid函数进行区域筛选,缩小了下一步目标搜索范围,并利用Haar小波方法从区域纹理入手,综合考虑了目标区域及其周边8方向邻域的相似性关系,明确了图像目标识别中对红外实时图目标的提取;最后运用目标模板的先验知识进行目标识别。实验结果表明,该算法的目标识别率相对于模板匹配算法和otsu阈值检测算法得到了极大提高,对于复杂地面红外图像目标的匹配识别具有一定应用价值。A new infrared target recognition algorithm based on image grayscale was presented to solve the problems of complex background, the impact of similar goals on target identification and low recognition rate. Firstly, a multi-level morphological target detection operator was designed and combined with region growing method to detect the target re- gion. Secondly, the introduction of sigrnoid function was used to screen the candidate target region, narrowing the searching range, and the Haar wavelet method was used to consider the similarity relations surrounding the target re- gion neighborhood of 8 directions starting from the region texture in order to confirm the target extraction in infrared- red real-time target recognition. Finally, the detected results were precisely matched using the prior of target template knowledge. Experimental results show that the algorithm has a higher target detection recognition and get great im- provement compared with template matching algorithm and otsu threshold detection algorithm. The new algorithm has a certain application value for infrared image target in complicated background.

关 键 词：目标识别 多级形态学 SIGMOID函数 HAAR小波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]