The Implementation Of Mobile Video Network Situation And Guidance Application System

The problem of Mixed Traffic is the important cause that leads to traffic accidents and operation efficiency dropping in the roads.And Hazardous Chemical Trucks will bring serious consequences in the event of accidents.Separating the Passenger and Freight Vehicles,strengthening the regulatory of Hazardous Chemical Trucks,and releasing the dynamic traffic states have important significance for the traffic’s safety and smoothness.The video detection method using the surveillance camera on the road has advantages of easy deployment and low cost.The thesis is based on the project of Jiangsu Provincial Science and Technology Department " The demonstration system based on the detection,mining,convergence and publishing of the freeway traffic sensor network information and assisting decision"(Project Number:BE2011822).The thesis described the research status of "Passenger and Freight Vehicles’Separating" and "the Detection of Hazardous Chemical Trucks";analyzed and compared different Vehicle Classification methods like Induction Coil,Laser Profiling,Image Analysis;improved the method which use the SIFT features and the SVM classifier via Sparse Coding;proposed a video based method for detecting the implementation of "Passenger and Freight Vehicles’ Separating" and a plan to recognize Hazardous Chemical Trucks;and built a visualized system releasing the dynamic traffic states.For the problem of Vehicle Classification,we analyzed and improved the traditional SIFT-SVM method,recoded the images’ SIFT features using Sparse Coding to capture structures and patterns inherent in the SIFT features,made the feature vectors sparse in high-dimension spaces to gain better Linear Classification performance.Eventually the comparative experiment proved that the SIFT-SC method is 10%higher than the SIFT-SVM method in accuracy.For the problem of separating the Passenger and Freight Vehicles and recognizing Hazardous Chemical Trucks,we applied the SIFT-SC method in practical projects.The system extracts the vehicle image blocks in the video using Gaussian Mixture Background Model computes the key points’ SIFT feature vectors,recodes the vectors through the Sparse Coding dictionary,and recognizes the vehicle type via S VM classifier.At the same time,the system distinguishes the driving lane by image block position and the lane line’s mapping relationship between image and 3D space.Combining the vehicle type and driving lane information,the vehicles violating related regulations will be detected.For the problem of releasing the dynamic traffic states,we designed a quick responsive and high-capacity data center which stored the real-time datum in relational database and transferred them to HDFS regularly,introduced the common methods for traffic data visualization,built a real-time traffic states visualized system using the WEBGIS,and realized many visual functions for history traffic datum.The thesis’ innovation points are as follows:1)It improved the SIFT-SVM vehicle classification method.The improved method recodes the images’ SIFT feature by Sparse Coding,classifies them using SVM.And it achieves a high accuracy for 90%.2)It proposed a video based detecting method for "Passenger and Freight Vehicles’ Separating".It adopts SIFT-SC method to recognize vehicle type,detected the vehicles violating related regulations considering the position of the lane lines and vehicles and has advantages in easy deployment,wide application,low cost.3)It designed a ITS data storage and analysis system which combines the relational database and HDFS together.It analyses the real-time data in SqlServer,and stores the original history data in HDFS satisfying the needs for high speed and large capacity at the same time.