Traffic Congession Detection through Modified Resnet50 and Prediction of Traffic using Clustering

Today metro cities in India are facing more problems due the traffic congestion in the roads that connects different neighboring cities. Due to overwilling of population and covid - 19 pandemic that increased the usage private vehicles by the mankind. Many of the existing methods use the sensor to detect the number of vehicles in that location and implied on to the google maps. The model has been proposed to overcome the issue by considering the existing video cameras been installed in different location in the signals and crowed places. The proposed model incorporated in the edge computing through Nvidia Jetson nano for the object detection using Common Objects in Context (COCO) to detect the boundary boxes for the ten categories of the object. The detection Score is obtained by considering probability of the detection boxes identified in the particular location with respect to time. The accuracy level of each object in the frame is detected and filtered based on the threshold value such that the false positive and true negative can be avoided in the Contributory matrix. Based on the results obtained from the model the location wise table that classify the images with respect low medium and High. The proposed model proves that accuracy of predicting the traffic based on three categories as Low, Medium and High are estimated correctly. The results are apprised based on the accuracy of the particular object identified in the image or frame with respect to number of images tested for the accuracy. The mean square errors are assessed based on the number of objects identified on each category in a particular image.