Visual detection method for road driving area

Abstract

The invention discloses a visual detection method for a road driving area. The method comprises: using superpixels as a middle layer feature perception consistency unit, based on middle layer superpixel segmentation, establishing an energy function, data dependence items of the energy function being defined by colors, textures, and position features of the superpixels, data interaction items introducing interaction of spatio-temporal neighborhood superpixels, and the data interaction items being defined by labels thereof and color feature differences; in addition, according to circulation of initializing classification labels, initial global energy calculation, local energy comparison, and global energy comparison, determining implementation energy minimization. The method can effectively detect road driving areas in images and videos, and the method is simple and effective.

Description

technical field [0001] The invention belongs to the fields of image processing, computer vision and pattern recognition, and in particular relates to a visual detection method of a road driving area. Background technique [0002] The detection of road driving areas has important applications in the fields of image processing, computer vision and pattern recognition. According to the detection results of road driving areas, the spatial range of vehicles and pedestrians in the video image scene can be determined, and it can serve the field of intelligent transportation systems ; The classic Markov random field method can realize pixel-level image region detection. The basic idea is to apply context constraints to adjacent elements in the image; in order to increase the feature dimension used for detection and discrimination and improve the speed of region detection, Wang method (refer to Wang's method: Wang XF, Zhang XP.A new localized superpixel Markov random field for image ...

AI Technical Summary

Problems solved by technology

[0002] The detection of road driving areas has important applications in the fields of image processing, computer vision and pattern recognition. According to the detection results of road driving areas, the spatial range of vehicles and pedestrians in the video image scene can be determined, and it can serve the field of intelligent transportation systems ; The classic Markov random field method can realize pixel-level image region detection. The basic idea is to apply context constraints to adjacent elements in the image; in order to increase the feature dimension used for detection and discrimination and improve the speed of region detection, Wang method (refer to Wang's method: Wang XF, Zhang XP.A new localized superpixel Markov random field for image segmentation[C].IEEE InternationalConference on Multimedia&Expo.2009) proposes an image segmentation based on local superpixel Markov random field method, using a cyclic iteration method to replace pixels with superpixels to realize image region detection, but this method has defects such as insufficient use of image time domain information, complex energy function iterations, and low efficiency; Pei method (refer to Pei's method: Pei SC ,Chang WW,Shen CT.Saliency detection using superpixel beliefpropagation[C].IEEE International Conference on Image Processing,2014.) proposed an image region detection algorithm based on superpixel saliency features, first segmenting the image into middle-level superpixels , and extract the visual features of a single superpixel, based on which a Markov random field algorithm is established to optimize the salient area of ​​the image. stronger

Images