Research On Road Detection Technology Based On Semi-random Forest Algorithm

Automatic driving which belonging to the field of artificial intelligence is a very strong practical technology.This technology allows the driver to liberate from the driving behavior in the long time,and effectively reduces the probability of traffic accidents,improves the efficiency of road traffic.At present,the main methods of road recognition can be divided into two directions:road boundary identification and road surface recognition.In this paper,an improved semi-random forest algorithm is proposed to realize the fast and accurate identification of the effective road surface from the complex road scenes.Firstly,the road image feature extraction is the important basic,after free combination of different features,we find out the highest weight of the two group characteristics;then use the SLIC algorithm to finish super pixel segmentation,and based on the experiments we find out the optimal parameter setting;then different classifiers are used in road recognition,the experiments were carried out,according to the Random Forest classifier,we put forward the corresponding improvement methods,with semi-random feature selection instead of the original random feature selection,we complete decision tree training;then through the exploration of the theory of the feasibility of the proposed algorithm,and the effect is verified through the experiment;finally,we do the error classification analysis.Based on the semi-random forest algorithm,we enhance the classification capability of single decision tree in both the accuracy and timeliness,this can be applied to real-time road recognition technology,and has a wide application prospect.