Research Of Road Markings Recognition Based On Driving Video And Its Android App Development

Transportation has always been an important issue in the development of human society.With the increasing number of motor vehicles,the problems of various traffic accidents are becoming more and more prominent.Large part of them are only caused by the human error of the drivers.In order to solve the above problems,we can develop driver assistant system of self-driving technology under the current technical conditions,which can greatly reduce or even prevent the occurrence of traffic accidents and thus improve road traffic efficiency.Driving videos contain rich traffic information.To extract useful information from the videos is of great significance for the intelligent driver assistant technology.The detection and recognition of road traffic markings is an important part of the intelligent driver assistant system.This paper focuses on the study of the road traffic markings.The purpose is to extract and recognize the road traffic markings from driving videos and make an Android application with it.In order to implement such a system,the main contribution of this paper includes:(1)For the segmentation and extraction of the road markings,we delve into the method of digital image processing,including preprocessing techniques,binarization methods,morphological filtering and contour finding algorithm.And we finally use a series of methods that can extract the target.(2)To choose the eigenvector which can effectively represent the target,two different features are studied: the HOG feature descriptor and the Hu invariant moment.The experiment shows that the HOG feature descriptor outperforms the Hu invariant moment and thus it is used as the feature of the system.In order to classify the features,a multi-class support vector machine(SVM)is developed and its classification results perform well.To improve the algorithm,the video frame is windowed and the shadow effect is eliminated.(3)Several techniques have been used to implement the development of the Android application.The overall design separates the interface and its implementation,that is,the separation of client and algorithm.We use the OpenCV library for the algorithm design with the JNI native method.The client is written in Java programming language.The test result shows that the Android application can well satisfy the need of use.