双目凝视系统中运动点目标实时检测算法

通过双目凝视系统视场重叠区进入系统的信息量大于通过非重叠区进入系统的信息量,根据这一特征,建立了应用于双目凝视系统的运动点目标并行实时检测算法。算法采用差分向量范数法对原始图像序列做预处理,去掉大量低频噪音和背景,采用光流场法从预处理后的图像中分离可疑运动点目标,再用提出的空间时间并行快速判定准则对分离的可疑运动点目标进行运动连续性和一致性判定,识别出真正的运动点目标。实验结果表明,建立的运动点目标检测算法的平均速度比单目视觉系统提高了50%,在图像信噪比不小于3dB时准确判定识别的概率为95%。

Real-time detection of moving point targets in staring binocular imaging system

The information quantity passed through the overlapped Field Of View (FOV) is larger thanthat passed through the non-overlapped FOV in staring binocular imaging system, a new parallelhigh-speed automatic detecting algorithm of moving point targets is established based on this feature.The detecting algorithm uses the Norm of the difference vectors of the processed images sequence as apreprocessing to get rid of most of low-frequency noise and background, then Optic Flow is used tosegment the doubtful moving point targets from the images remained after preprocessing, and finally thenovel space-time parallel high-speed determining method proposed in this paper is used to determinewhether the doubtful moving point are true targets or not. The experiment results show that with theestablished algorithm, the average determining speed for moving point targets is raised 50% than that ofthe monocular imaging system, and the correct determining probability is 95% when the SNR ofprocessed images is no less than 3 dB.