摘要
对红外序列图像小目标检测过程中出现的误差做出了分析 .在检测过程中采用形态学中的Top hat滤波 ,并根据目标运动的连续性去除噪声和云团 .对目标与背景灰度均值之差、目标灰度方差和目标面积这三个特征量进行了研究 ,发现它们保持相对的稳定 ,可以作为小目标的不变特性 ,利用这些特征量设计了RBF神经网络并对检测结果进行评估 。
The detection method of infrared small target is summarized. The errors that appear in detecting course are analyzed. In detecting course, morphologic operator that is Top hat operator is adopted. According to the continuity of target, noise and cloud group can be excluded. Three features are namely presented the difference between target gray mean and background gray mean, gray variance of the target and the target area to measure the results of detection. These targets′ invariant characters can be regarded as evaluation criterion verified by the results of experiments.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2002年第3期83-85,共3页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国防科技大学ATR重点实验室基金项目 (0 0JS93.2 .1.JW 0 5 15 ) .
