加速概率存在场目标检测的改进算法-计算机工程与应用.PDFVIP

146 2015 ，51（14） Computer Engineering and Applications 计算机工程与应用 加速概率存在场目标检测的改进算法 尹 辉，孙水发，罗 恒，雷 林，董方敏 YIN Hui, SUN Shuifa, LUO Heng, LEI Lin, DONG Fangmin 1.三峡大学 计算机与信息学院，湖北 宜昌 443002 2.三峡大学 湖北省水电工程智能视觉监测重点实验室，湖北 宜昌 443002 1.College of Computer and Information Technology, China Three Gorges University, Yichang, Hubei 443002, China 2.Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering, China Three Gorges University, Yichang, Hubei 443002, China YIN Hui, SUN Shuifa, LUO Heng, et al. Improved algorithm to accelerate target dectection of probabilistic occu- pancy map. Computer Engineering and Applications, 2015, 51 （14）：146-150. Abstract ：Probabilistic occupancy map is an algorithm to detect targets on multiple views, it estimates the probability that target appears on each 3D position at the ground of detection scene based on the binary foreground images shot on each view. When the area of detection scene is large and therefore many 3D positions are to be detected, it will be very time-consuming to estimate the probability that target appears on each 3D position. A method to accelerate the POM algo- rithm is proposed in this paper ：the number of 3D positions that POM algorithm needs to process is reduced substantially based on the binary foreground image and the corresponding 3D positions. The experimental results show that the speed of the algorithm can significantly improve with little reduction of detection accuracy when there is only one target or mul- tiple targets appear in the scene appear. Key words ：Probabilistic Occupancy Map （POM ）; multiple views; target dectection; 3D position; speed 摘 要：概率存在场（POM ）算法是一种多视角目标检测算法，其根据各视角对应的二值前景图像，估计检测场景中 所有3D 位置对应地面上存在目标的概率。当检测场地面积过大时，POM 算法估计所有3D 位置上存在目标的概率 会十分耗时，拖累算法速度。提出了一种加快POM 算法速度的方法：根据二值前景图像和3D 位置信息，大幅减少 POM 算法需要处理的3D 位置数量，从而提高算法速度。实验结果表明：在对算法检测结果影响很小的情况下，对于 单目标和多目标检测场景，算法速度都得到了较大的提升。 关键词：概率存在场算法；多视角；目标检测；3D 位置；速度 文献标志码：A