Research On Efficient Mark Recognition Technology Based On Perspective Deformation Reduction Transform

In the wake of the rapid development of society and road traffic system,more and more people have their own car in China.However many troublesome questions are accompanying,such as traffic jams and traffic safety.The intelligent transportation system which aims to solve those traffic problems gets more attention now,and governments have invested a lot of resources to study it in the world.As a key part of intelligent transportation system,the intelligent vehicle is an intelligent platform which integrates multiple functions,such as environment perception,decision-making,and driver assistance,etc.Intelligent vehicle represents the future development direction of the car,and has broad application prospect.As the key component of environment perception system of intelligent vehicle,the traffic signs detection and recognition technology based on computer vision has been the focus of research topics for scholars.The technology detects and recognizes the traffic signs on the road in time by extracting the characteristics of the traffic signs on the road images collected by the camera.However,the image taken by the camera has a strong perspective deformation,which will cause distortion of the traffic signs.This distortion is not conducive to subsequent detection and recognition.Therefore,it is a key point for intelligent vehicle to weaken or even eliminate perspective deformation on road images.Based on this,this paper proposes a method,which can rapidly weaken the perspective deformation of road images.This method can reconstruct the images captured by the camera in other lanes so as to weaken the deformation of traffic signs in each lane and use the processed images for the detection and recognition of traffic signs.In addition,in view of the fact that our method can also extend the image and increase the original image data volume,this paper also uses the target detection based on deep learning to train the extended image and detect the traffic signs on the original image.First of all,this paper introduces the background and significance of rapidly weakening perspective deformation on the road image,and then then analyzes the problems faced by traditional detection methods and target detection methods based on deep learning,thus citing the main work and content arrangement of this paper.Secondly,this paper gives a brief introduction to the camera imaging model,lists the coordinate systems involved,analyzes how the points in the three-dimensional world map to the two-dimensional plane,and describes the inverse perspective mapping based on the simplified camera model.Next,this paper proposes an effective algorithm to weaken the perspective deformation of traffic signs.For the road images captured by the camera,we first preprocess them in order to remove some useless information(such as the sky),and then rebuild the images by changing the viewpoint as if captured by the camera located in front of each lane.In order to speed up the subsequent detection and recognition of traffic signs,we also determine a rough detection range for each lane,and apply the processed images for the detection and recognition of traffic signs.Finally,taking into account that the proposed algorithm can reconstruct the road images when the camera is located at the front,left and right direction according to the original image.This paper extends the road image database and apply the current mainstream target detection based on deep learning to detect traffic signs on images.