Markowitz投资组合模型改进的三帧差分法运动目标检测

Moving object detection method based on three-frame difference and improved by Markowitz portfolio model

移动机器人的目标检测要求其对特定的静止或运动物体进行运动分析及检测。以Voyager-III移动机器人系统为研究对象,实现非理想光照下,对橘红色目标足球的运动检测。提出在传统三帧差分法基础上,先利用Markowitz投资组合模型进行足球目标的特征提取,将场地非感兴趣的目标中,出现全部像素值发生变化的目标去除,再进行图像帧间差分。利用CCD摄像机对比赛环境中足球的运动轨迹进行录制,选取具有代表性的各帧视频图像、Markowitz算法优化后的差分图像和跟踪图像,结果表明跟踪图像不含非目标物的干扰,克服了差分图像存在空洞的问题,为移动机器人提供了一种实用的运动目标检测方法。

The motion analysis and detection on the static or moving object should be made for the mobile robot involved in detection work. Taking Voyager-III mobile robot system as the investigated subject, the detection of orange soccer in motion is accomplished under the un-ideal illumination condition. The improved approach based on traditional three frame time difference is proposed. The Markowitz portfolio model is introduced to realize the feature extraction of the target soccer in the first step, which helps to remove the objectives of non-interest, so that it will not puzzle us with the varied total pixel value. In sequence, the results are inserted to the next procedure of three frame time difference. The motion track of the soccer, under the real competition environment, is recorded by a CCD video camera, and the final results, including the video images, optimized difference images, and tracking images of representative frames are presented. The final results demonstrate that the interferences of other objects are not involved in the tracking images. Besides, the weak status of difference images with holes is improved, which indicates that the method above adapts to practical application for mobile robots in moving object detection area.