Vehicle Target Tracking Based On Layered CNN In Big Data Environment

With the rapid development of smart city and intelligent transportation,which generate massive traffic surveillance video data.How to effectively track these vehicle targets from video data has become an urgent problem at present.Traditional target tracking methods have shortcomings in precision and application scope,which cannot meet new demands while the development of deep learning shows a new solution.At the same time,the traditional single machine method cannot handle massive vehicle video data that are supposed to be processed efficiently and in time,thus big data processing framework is chose as the new approach.In the above background,this paper combined deep learning with vehicle target tracking,then proposed a novel layered convolutional neural network(LCNN),and also designed and implemented a Vehicle Target Tracking based on Layered CNN algorithm.This novel LCNN divides traditional CNN into two layers: one layer is used to learn the general feature of the vehicle targets in all tracking sequences,and the other layer is to learn the unique feature of each tracking sequence.This method solves the problem of existing target tracking algorithm such as low accuracy and loss of tracking target,and improves the accuracy of vehicle target tracking,which is verified in comparison analysis with other existing target tracking methods.The proposed algorithm in this paper has a low efficient on single machine,so this paper combined this algorithm with Spark,designed and implemented parallelized VTT-LCNN algorithm on Spark.Based on this algorithm,analyzed,designed and implemented a vehicle target tracking framework in big data environment from perspective of software engineering.After test on small self-built Spark cloud platform,proved that this parallelized VTT-LCNN algorithm and vehicle target tracking framework is feasible.This results also verified that the parallelized VTT-LCNN algorithm can effectively complete the vehicle target tracking task,and has good performance and scalability on Spark cloud platform.