论文部分内容阅读
本文以多帧检测成功为条件,通过预测目标在下一单帧上可能的存在区域,再在此区域内进行目标的Meyman- Pearson检测。由于信噪比很低,区域内可能会出现很多虚假目标,利用已获知的目标初始信息和数据关联技术,确定其中某一个为真实目标成为关键决策问题。为此,本文将概率数据关联技术应用到图像序列中目标检测领域,得到了重要的理论分析和实验结果。高分辨检测器接收已获知的目标初始信息(目标的初始位置、运动速率以及目标亮度)后,根据概率数据关联技术进行正确测量值的确认,再用卡尔曼滤波器来预测目标在下一帧的可能状态。文中还给出了理论分析、公式推导过程和实真结果。
In this paper, the successful detection of multi-frame conditions, by predicting the target in the next single frame may exist in the area, and then in this area for the target Meyman-Pearson test. Due to the low signal-to-noise ratio, many false targets may appear in the region. Using known target initial information and data correlation techniques, determining that one of them is a true target becomes a key decision issue. Therefore, this paper applies the probabilistic data association technique to the field of object detection in image sequence, and obtains important theoretical analysis and experimental results. After the high-resolution detector receives the initial information of the target (the initial position, the moving velocity and the target brightness) of the target, the correct measurement value is confirmed according to the probabilistic data correlation technique, and then the Kalman filter is used to predict the target in the next frame Possible state The article also gives the theoretical analysis, formula derivation process and real results.