一种可见光成像的目标检测跟踪方法

A detection and tracking algorithm for optical target

在末制导成像阶段目标占据视场的大部分面积,形状信息丰富,这时对目标体上某一固定点的稳定跟踪成为最主要的需求。针对图像灰度过于集中目标难于提取的问题,在图像预处理阶段提出了依据概率统计的图像灰度均衡的方法;为解决图像边角点的快速检测问题,提出了图像二维投影和滑动卷积相结合的方法;最终为实现快速稳定的跟踪,提出了依据角点匹配的形心跟踪方法。根据此算法的特点研制了FPGA+DSP的硬件平台,充分利用FPGA的并行流水特性和DSP的逻辑运算特点,能快速检测到目标形心和进行稳定地跟踪,验证了方法的有效性和实用性。此方法应用在某课题上,跟踪效果良好,为试验的圆满成功提供了保障。

In the terminal guidance phase imaging target occupies most of the field and shape information is rich.The main requirement is to stable tracking one part of the target.The gray scale of original image is too intensive and target is difficult to detect,so a method of gray-scale image histogram equalization is proposed.To fast detect corners of the image,a method of two-dimension projection and slide convolution is proposed.At last to reach real-time and stable tracing,so a type-heart tracing method based on corner match is used.And basis the feature of this algorithm,a hardware platform based on FPGA and DSP is designed.FPGA is good at parallel pipeline and DSP is good at logical operation.The algorithm is used by one item,and the target tracing is favorable.The algorithm ensures this item is successful.