Research On Method Of Markov Random Field Road Scene

Road scene segmentation is a crucial part of robot environmental understanding. For different scenes, traditional algorithms may bring out incorrect segmentation results. It is an important problem inroad scene segmentation to reduce the rate of incorrect segmentation.The focus of this paper is to extract the road from road scene which is road scene segmentatioa The main work is as follows:(1)First, this paper studies Markov Random Field (MRF) and the related theories, and applies the theories to road scene segmentation. Also, this paper applies ICM and simulated annealing method for the energy optimization in MRF to cut up the road scene.(2)Then, this paper studies Markov Random Field (MRF) and grayscale symbiotic moment, and applies this to improve the energy function. Calculate the energy of road scene by MRF, and improve this energy by modifying the potential function. Research the basic framework and principal of the Particle Swarm Optimization (PSO), apply this to improve the energy function, this can help to cut up the road scene.(3)Finally, this paper studies road scene segmentation via Graph Cut. Graph cut is an algorithm to cut up the road scene combined with color and texture. Initialize the road scene by K-means, the initialization result can be used by Graph Cut, use the result of this frame to initialize the next frame and cut up the next frame. Loop all of these steps.