【作者】 李东光； 殷俊； 房慧敏；

【Author】 LI Dong-guang, YIN Jun, FANG Huknin(School of Mechatronics, Beijing Institute of Technology, Beijing 100081, China)

【机构】 北京理工大学机电工程学院；

【摘要】 昆虫复眼的相关型初级运动监测器(EMD)模型仅仅是根据物体的图像强度来检测运动物体的,完全不需要进行物体识别,因而特别适用于检测伪装的运动目标。相关型EMD模型对运动方向极为敏感,可由输出估算图像的运动速度。当运动速度与延迟时间匹配时,EMD模型的输出为最大。单个EMD对的输出是局域的,可能具有不确定性。因此,实现大视野运动检测需要一定规模的二维EMD阵列对EMD的局域输出进行空间整合。由此设计的计算机仿真系统能检测出具有一定背景干扰的运动物体,验证了该仿生模型在工程上应用的可行性.对今后进一步研究仿生探测技术具有一定的意义。更多还原

【Abstract】 The correlative Elementary Motion Detector (EMD) model relied merely on the object intensity without the need to identify objects. It worked very well even when the moving objects were completely camouflaged in visual scenes. The correlative EMD model was very sensitive to motion direction. Velocity could be also estimated by the model output. When the speed of the moving object matched the delay time used in the model the output of the model was maximum. However, the output of a single pair of EMD was local and might be ambiguous. Therefore, wide-field motion detection required a 2D array of EMDs integrating in parallel the local signals over the EMD array space. Then according to this model, the computer vision detection system can detect the moving object in the certain interfered background. The feasibility of imitating biologic model is fully demonstrated in the engineering application, and it absolutely lays a solid foundation for the future study of detection technology of imitating biologic vision.更多还原