Generic obstacle detection on roads by dynamic programming for remapped stereo images to an overhead view

Abstract

This uaver Drouoses an avuroach to extract , . . . .. 3) Extraction of matching points between stereo images 4) Segmentation of obstacles generic obstacles on roads using remapped stereo images to an overhead view. The proposed approach uses the fact that the information of roadsurface on the remapped imane is distorted by an obstacle, and formulates the generic obstacle detection (GOD) problem (IS a dynamic pronramminn (DP), which contributes to search c for 5 ) Depth computation . --. corresponding pea.& on polar histograms constructed by vertical edge components of the remapped stereo images. The corresponding p e a k determine the estimates of obstacles'positions. The approach has features that it is not largely affected by intensity diyerence between apair of stereo images and does not depend on the typical stereo matching, and identifies the obstacle's position quite well.

A stereo matching algorithm based on top-view transformation and dynamic programming for road-vehicle detection

“…In this paper, the uniqueness constraint, ordering constraint and disparity-smoothness constraint are considered in a cooperative fashion on the dynamic programming [16]. In various stereo vision applications, the complex epipolar geometry [3] is necessary for epipolar line computation to reduce search space or image rectification when adjusting both images in parallel.…”

“…To minimize stereo matching errors caused by these difficulties, several constraints such as epipolar constraint, uniqueness constraint, ordering constraint and disparity-smoothness constraint are imposed [2]. Recently, a great deal of research has been performed on stereo matching using stochastic and optimization methods such as gradient-based optimization [20], nonlinear diffusion [21], dynamic programming [16,18,19], genetic algorithm [22,23], neural network [24,25], belief propagation [26], intrinsic curves [27] and graph cuts [28]. Features for matching are also expanding from pixel, block and simple features such as a vertex, edge, line and corner to local descriptors of an image [29,30].…”

TerraMax™: Team Oshkosh urban robot

“…Obstacle detection is performed in two steps: first the two images, acquired simultaneously, are preprocessed in order to remove the high lens distortion and perspective effect, a thresholded difference image is generated and labeled (Bertozzi, Broggi, Medici, Porta, & Sjögren, 2006), and then a polar histogram-based approach is used to isolate the labels corresponding to obstacles (Bertozzi & Broggi, 1998;Lee & Lee, 2004). Data from the LIDARs are clusterized so that laser reflections in a particular area can boost the score associated with the corresponding image regions, thus enhancing the detection of obstacles.…”

Clustering of Stereo Matching Data for Vehicle Segmentation

“…후보점 추출 본 장에서는 참고문헌 [3,7]의 후보점 검출 알고리즘의 개 요를 설명한다. 그 이유는 후보점 클러스터링에서 사용할 후 보점의 속성이 어떻게 형성되었는지 보이기 위해서다.…”