Learning to segment roads for traffic analysis in urban images

摘要: Road segmentation plays an important role in many computer vision applications, either for in-vehicle perception or traffic surveillance. In camera-equipped vehicles, road detection methods are being developed advanced driver assistance, lane departure, and aerial incident detection, just to cite a few. surveillance, segmenting information brings special benefits: automatically wrap regions of analysis (consequently, speeding up flow videos), help with the driving violations (to improve contextual videos traffic), so forth. Methods techniques can be used interchangeably both types application. Particularly, we interested from remaining image, aiming support tasks. our proposed method, relies on superpixel based novel edge density estimation method; each superpixel, priors extracted features gray-amount, texture homogeneity, motion horizon line. A feature vector all those feeds machine classifier, which ultimately takes superpixel-wise decision not. dataset challenging scenes was gathered video surveillance cameras, city, demonstrate effectiveness method.