摘要
形心跟踪算法因为简单、执行速度快而得到广泛采用。然而常规形心跟踪算法因激光雷达图像受到散斑噪声影响而变得不适合。改进算法结合了图像的滤波处理,提高了适用范围。但是这种基于图像灰度均值分割的形心跟踪算法对于小目标图像都会跟踪失败。分析了大目标图像和小目标图像的直方图,指出其均由一个背景的尖峰和目标的拖尾组成,理想的分割阈值点应为尖峰和拖尾的临界点。据此提出了基于累积直方图的形心跟踪算法。此算法以图像的累积直方图拐点为分割阈值,不受目标区面积大小的影响。处理结果表明,对于大目标图像和小目标图像均有较好的效果,并且保持了常规形心跟踪算法执行速度快的特点。
Centroid tracking algorithm is widely used because of its simplicity and quick run speed. The original centroid tracking algorithm is no longer suitable because of the speckle noise in ladar image. The enhanced algorithm with the image ftlter processing extends its application range. However, the centroid tracking algorithm based on the image gray-mean segmentation fails when tracking small target images. The histograms of big target and small target images are analyzed. It is pointed out that they are made up of peaks from background and tails from target.And the ideal segmenting threshold should be at the point of the critical point between the peak and tails. Centroid tracking algorithm based on cumulative histogram is presented,which selects the knee point between steep rise and slow rise in the cumulative histogram of the image and is independent of the target area size. The processing results show that the algorithm performs well in big and small target images,and maintains the advantage of quick processing speed of original centroid algorithm.
出处
《红外与激光工程》
EI
CSCD
北大核心
2006年第6期696-699,共4页
Infrared and Laser Engineering
关键词
图像处理
激光成像雷达
目标跟踪
形心算法
累积直方图
Image processing
Imaging laser radar
Target tracking
Centroid algorithms
Cumulative histogram,
