Research of Robust Algorithm for Understanding a Situation in a Dynamic Environment

1.针对光照变化和阴影的鲁棒目标检测算法“平面投影立体视法”能够有效消除运动物体投射的阴影区域。然而，该方法存在的主要问题是，它不仅检测到移动的物体，而且还检测到一张图像中被物体遮挡的道路区域。在本研究中，通过设计新的算法来消除遮挡区域，将平面投影立体视法检测到的区域的轮廓与原始图像的边缘进行比较，从而解决了这一问题。恶劣天气下的鲁棒运动目标检测算法在能见度较差的天气中，如降雪或浓雾天气，能见度在短时间内变化较快，每帧中每个像素的强度变化较大。为了克服这些问题，该算法将输入图像划分为网格区域，并在每个区域计算两个累积帧数不同的直方图之间的相互关系。一个短的累积直方图，由积累少量帧数生成，当移动物体进入该区域时，它会迅速变化。另一方面，由于积累了更多帧数而生成的长累积直方图变化缓慢。因此，通过测量短累积直方图和长累积直方图之间相互关系的变化来检测运动目标。从图像序列中测量能见度的算法该算法将输入图像划分为网格区域，通过测量每个区域的强度方差变化来解决降雪或黑暗的程度。基于立体图像序列的手势识别算法提出了一种扩展时间模板的手势识别方法，使它们不仅可以包含垂直和水平运动，还可以包含双目立体视觉获得的深度信息。

1.Robust object detection algorithm toward sunshine change and a shadow"Planar Projection Stereopsis Method" is effective to eliminate shadow areas casted by moving objects. However, this method has a major problem that it detects not only moving objects but also road areas occluded by objects in one image. In this research, this problem has been solved by conceiving new algorithm for eliminating occluded areas, comparing contours of areas detected by the planar projection stereopsis method with edges in an original image.2.Robust moving object detection algorithm in bad weatherIn the weather whose visibility is bad, such as in a snowfall or in a dense fog, the visibility changes rapidly in a short time and the intensity of each pixel changes hard every frame. In order to overcome these problems, the proposed algorithm divides an input image into grid regions and in each region, calculates a cross correlation between two histograms whose accumulated number of frames are different. A short accumulated histogram, generated from accumulating a few number of frames, changes quickly whenever moving objects go into the region. On the other hand, a long accumulated histogram, generated from accumulating the more number of frames, changes slowly. Therefore, moving objects are detected by measuring a variation on a cross correlation between a short accumulated histogram and a long accumulated histogram.3.Algorithm for measuring visibility from image sequencesThe proposed algorithm divides an input image into grid regions and solves the degree of the snowfall or darkness by measuring the change of intensity variance in each region.4.Gesture recognition algorithm using stereo image sequencesA gesture recognition method extending Temporal Templates so that they can contain not only vertical and horizontal motion but also depth information obtained from a binocular stereopsis has been conceived.