Research On Integrated Road Recognition Method Based On Spatial Line Model For Infrared And Visible Light

The development of society leads to the start of intelligent vehicle research during the eighties of the last century.The purpose of the intelligent vehicle is to improve the safety and reliability of driving by using the machine rather than people.Road detection is the basis and core of intelligent vehicle.The vision-based road detection method has been investigated extensively.Because of the rapid development of imaging equipment and the characteristics of the sensor itself:imaging equipment can provide more information than other sensors.Furthermore,it is similar to the way people perceive the environment.Therefore,vision-based road detection is widely used in intelligent vehicle systems and develops quickly.However,road detection is a challenging task,changes in lighting and weather,roads with different shapes or size,as well as the presence of different objects(vehicles,pedestrians,and infrastructure elements)have brought great challenges to road detection.In this paper,the road region detection algorithm based on multi-sensor fusion is investigated on the basis of lane detection,road region detection and road boundary detection,so as to improve the robustness and accuracy of vision based road detection in different environments.In this paper,we use different road classifiers to extract road region,namely:road region classifier based on shortest path,road edge classifier based on Prewitt and lane classifier based on gradient.Road region detection method based on the shortest path can distinguish the non-road region and road region with similar color information.At the same time,this method can be effectively applied to complex road scenes with good adaptability to the change of road shape.Since the SPRAY model is not robust to the change of the road width,this paper proposes a road region detection method based on binary spatial ray feature.The method weakens the influence of distance on the feature,and it works better in the stability of the road region feature even if the width of the road is changed.In addition,in order to accelerate the process of SPRAY feature extraction,we use the inter-frame relation to reduce time.The similarity between two adjacent image can be used to avoid redundant feature extraction and prediction.Based on SPRAY model,a new road detection method for the fusion of visible light and infrared images is proposed to solve the weakness that visible light camera cannot work during the night.The infrared and visible light images of the same road scene are captured.Then merging the two images to obtain the final road detection result such that the method can do work around the clock.In order to solve high time complexity of the fusion of infrared and visible light images road detection based on SPRAY model,this paper also proposes a method which combines random walk with SPRAY feature.This method preserves the SPRAY feature with good recognition ability,and avoids the complex SPRAY feature extraction for a large number of pixels.The core of the method is middle-line extraction of the road based on SPRAY feature,through the combination of SPRAY feature and machine learning to identify the middle-line,and on the basis of the middle-line with random walk algorithm to achieve the road region detection.