摘要
红外目标识别是影响精确制导武器命中精度的重要因素。针对因基准图与实时图存在差异导致模板匹配性能下降的问题,提出了一种基于边缘匹配的前视红外(FLIR)目标识别算法。在模板匹配的基础上,利用目标的邻域信息排除虚警,计算基于匹配代价矩阵的匹配置信度对候选匹配点进行排序,利用前面连续多帧图像的识别结果为当前帧识别提供信息,从而提高目标识别的准确性和鲁棒性。在2组不同目标类型的红外视频序列上所做的实验表明,该算法具有较好的识别性能。
Forward looking infrared（FLIR）target recognition is an important factor that influences the hitting accuracy of precision guided weapons.A forward looking infrared target recognition algorithm based on edge matching is proposed for the decrease in mask matching performance due to the dissimilarity between base image and real-time image.For each candidate target chosen by mask matching,the neighborhood information is used to eliminate the false alarms,the matching confidence is computed by using a matching cost matrix,and the recognition results of continuous foregoing frames are utilized to provide information for current frame recognition,thereby improving the accuracy and robustness of target recognition.The experimental results of two FLIR videos of different target types show that the proposed algorithm has an excellent performance.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2012年第3期271-277,共7页
Acta Armamentarii
基金
中国博士后科学基金项目(20100471768)
国家自然科学基金项目(61072141)
关键词
信息处理技术
景像匹配
边缘匹配
距离变换
匹配置信度
information processing
scene matching
edge matching
distance transform
matching confidence
