Research On Intelligent Transportation System Application Based On Flask And Machine Learning

The Intelligent Traffic System(ITS)is the development trend of urban transportation in the future.It can greatly solve problems such as traffic congestion and traffic safety,and improve the utilization rate of transportation.Intelligent transportation system control platform and video recognition are the most important functions in intelligent transportation systems.At the same time,modern communication methods are mainly based on network communication,and IPv4 addresses are seriously deficient.The next-generation Internet communication method based on IPv6 is an inevitable trend.This paper studies the latest Flask microframework and machine learning algorithms.And an intelligent transportation system control platform based on B/S mode is built efficiently on the Raspberry Pi.At present,the platform has realized remote control of an intersection signal.At the same time,it can remotely monitor the traffic conditions at various intersections and perform vehicle detection and attribute identification on vehicles in the video.The use of the Flask framework can map IPv4 addresses to IPv6 addresses,enabling network communications based on the IPv6 protocol.The construction of the intelligent transportation system control platform is mainly based on the micro-framework Flask under Python to complete the development of the front-end and back-end.This method has a short development cycle and low cost.At the same time,the platform is developed using B/S mode,which improves the efficiency of post-maintenance and management.The vehicle detection and attribute recognition algorithm based on machine learning is studied.Three models of SSD,Faster R-CNN_resnet101 and Faster R-CNN_nas are implemented respectively.The SSD model is a regression-based detection algorithm,and the Faster R-CNN model is a detection algorithm based on the Regional Proposal Network(RPN).Compared with the other two models,the recognition speed of the SSD model is better,but the average accuracy has decreased.The difference between Faster R-CNN_resnet101 and Faster R-CNN_nas is that the convolutional neural network has a different structure and the average accuracy is much higher than SSD,but the recognition speed is slow.And the objects identified by the three models are bounding boxes and there is a certain background,and the bounding box cannot be aligned with the target.Aiming at the above problems,an improved Faster R-CNN algorithm is proposed.A full convolutional network is added to the layer above the CNN feature mapping layer of the traditional Faster R-CNN to generate a mask,thereby improving the average recognition accuracy.The generated mask is aligned with the target.And the vehicle types are classified and trained according to the improved algorithm to realize vehicle attribute recognition.The experimental results show that the control platform of intelligent transportation system based on Flask framework is stable in performance and the various functions of the design are perfect.The control scheme can be switched freely.The comment function can realize normal interaction,and the video monitor function can realize monitoring and vehicle recognition.The platform can operate normally under the IPv6 environment.And the improved vehicle detection and attribute recognition algorithms have improved the average accuracy and can realize vehicle attribute recognition.