基于非递归迭代投影的实时深空目标检测算法

针对深空目标信噪比低、干扰多、传统方法无法检测准确性低的问题,提出一种基于非递归迭代投影的快速实时深空目标检测算法.该算法采用由粗到细的检测流程,首先根据图像特征自适应确定迭代投影次数提取所有疑似目标,然后通过邻域判决对目标进行轨迹关联剔除虚假目标,从而实现目标的准确检测.此外,为了提高算法的实时性,针对FPGA并行流水的特点,利用非递归编程架构对算法进行优化,提高了算法的并行度和运行效率.在SOPC硬件平台上的实验表明,该算法能够准确实时检测深空目标,对悬浮实验的圆满成功起到了决定的作用. 

Deep-Space Object Detection Based on Non-Recursion Iterative Projection

Traditional method could not detect all objects, when the deep-space image is low SNR and full of distortion. To tackle this problem, a fast and real-time deep-space object detection approach was proposed, which was based upon modified iterative projection and non-recursion programming method. In this algorithm, a coarse-to-fine detection process was used. Firstly, the number of projection was determined adaptively in accordance with the image unique properties to detect all objects. Then the trajectory matching was undergone by neighborhood decision approach to achieve accurate objects detection. To implement the algorithm in real time, a non-recursion program method was used to optimize the structure of the hardware. So, the proposed algorithm was easy to implement by FPGA. Experimental results on SOPC hardware platforms show that it can perform real-time detection accurately for the deep-space objects, and it plays a decisive role for the success of the experiment on suspension.