用于稀疏结构光视觉系统的传感器规划策略

Sensor planning strategy applied to sparse structured light vision systems

本文提出一种采用稀疏结构光实现对物体识别定位的方法。采用这种方法的关键是提高传感器规划策略的效率，为此作者提出了评价传感器方位的最大期望假设递减率概念，按照这种概念选择传感器采集信息的方位，可以使完成物体识别定位过程所需的平均调动传感器次数降为最低。为了使在线识别定位过程中的计算量降低，计算最佳视点的工作可以放在离线建模与仿真阶段进行，从而使在线传感器规划通过查找表及相应变换实现。本文着重说明这种新概念及其实现方法，并显示初步实验结果。

he authors propose a method for object recognition and localization using sparse structured light images.The key to this kind of methods is to raise efficiency of sensor planning. Here a concept of the maximum expected rate of hypothesis reduction(MERHR) is introduced to evaluate sensor positions. It is shown that the number of sensor placements in the process of object recognition and localization will be reduced to its minimum by adjusting the sensor positions according to the concept. In order to reduce computation in the on-line phase, the computation for optimal sensor positions is carried out off-line and verified with simulation.This makes it possible to accomplish sensor planning by just refering to a lockup table and performing a related transform. The proposed concept and its implementation are focused on, and the preliminary experimental results are also given.