旋转复杂背景中红外运动小目标实时检测

发布时间：2019-03-12 12:48

【摘要】：针对旋转复杂背景中红外运动小目标检测误检率高、实时性差等问题,提出了目标检测新算法。首先对图像进行中值滤波预处理,计算图像光流场,提取特征点,估算背景光流;然后设置阈值,判断提取备选目标特征点集合;最后通过特征点光流矢量角度、目标灰度值区间、目标特征点区域边缘检测的方法,排除备选目标特征点集合中的背景特征点,实时准确检测旋转复杂背景中红外运动小目标。实验结果表明,该算法能够准确地检测出红外多个运动小目标,检测率93.8%,平均虚警率0.126次/帧,平均每帧耗时15.53 ms,每帧图像处理的最大时间为20.45 ms,能够满足运动目标检测对实时性的要求。
[Abstract]:Aiming at the problems of high false detection rate and poor real-time performance of infrared moving small target detection in rotating complex background, a new target detection algorithm is proposed. Firstly, the image is pre-processed by median filtering, the image optical flow field is calculated, the feature points are extracted, and the background light flow is estimated, and then the threshold is set to judge and extract the set of candidate target feature points. Finally, through the feature point optical flow vector angle, the gray range of the target and the edge detection method of the target feature point region, the background feature points in the set of alternative target feature points are excluded, and the small infrared moving targets in the rotating complex background are detected in real time and accurately. The experimental results show that the algorithm can accurately detect multiple small infrared moving targets, the detection rate is 93.8%, the average false alarm rate is 0.126 times per frame, and the maximum processing time per frame is 15.53 ms, per frame. The maximum time of image processing per frame is 20.45 ms,. It can meet the real-time requirement of moving object detection.
【作者单位】：南京航空航天大学航天学院;光电控制技术重点实验室;
【基金】：国家自然科学基金(61471194) 航空电子系统综合技术重点实验室和航空科学基金联合资助(20155552050) 国家留学基金(201506835020)
【分类号】：TP391.41

【相似文献】