The Research On Road Object Recognition Based On Driving Video And Its Android App Development

With the rapid development of artificial intelligence and deep learning technology,intelligent assisted driving technology and driverless technology have become hot topics in the academia and industry.Road object detection and recognition is also the main content of the research.Therefore,robust road object detection and recognition algorithm plays an important role in the field of smart driving.This paper is mainly based on the research of road object detection and recognition algorithm in driving video.At the same time,based on the machine learning algorithm,we compose an Android application,and further the deep learning method is used to accurately recognize the road object.The driving video contains various types of road objects to facilitate the research of road object detection and recognition algorithm in this paper.In order to design such a road object recognition system,the main work of this paper is as follows:(1)For the problem of road object detection in driving video,the image processing techniques used in the paper are mainly including image graying,image normalization and histogram equalization.In addition,this paper mainly to use the cascade classifier and Haarlike features to detect the object in the preprocessed images for the object detection,and after that extracts the Histogram of Oriented Gradient feature and the Bags of Word feature from the detected object in the image.(2)For the multi-object classification and recognition of road in driving video,the paper mainly uses Support Vector Machine multiple classifier to recognize the detected objects.The paper mainly classifies the road objects into car,truck,minibus,person,cyclist and bus.(3)In the implementation of Android application,the paper uses C++ language and OpenCV computer vision library to write C++ program.After the C++ program is written,the paper uses the Java language and Java Native Interface programming technology to complete the writing of the Android application.Finally,we can detect and recognize the road objects in real time.(4)In view of the limitations of the traditional machine learning methods in real-time and accuracy,the paper adopts a deep learning method to accurately recognize the road objects in the driving video.The paper mainly uses the object detection framework based on Convolutional Neural Network.For the feature extraction layer,the convolutional layer of the residual network is used in our detection framework instead.A large number of experimental results show that the method based on the residual network has higher accuracy than that based on original network.Through the tests of these algorithms,it shows that the Android application can accurately recognize the road objects in real time,and the system based on deep learning methods performance better than that based on traditional machine learning methods in accuracy and real-time.