ANN-based traffic planning model using adaptive YOLOv5-aided object counting from surveillance videos

Recently, several researchers in the computer vision and multimedia fields have focused on traffic volume estimation and vehicle counting tasks by utilizing traffic videos as traffic has continued to increase due to population growth. To determine the traffic in an area for an intelligent transportation system, it is necessary to assess the volume of vehicles on the road. The widespread use of cameras in urban transportation systems has led to the centralization of surveillance footage as a source of data for performing various crucial activities. Additionally, big data evaluation and readily accessible mobile cameras have contributed to the recent rise in the popularity of real-time traffic management systems. However, finding a way to combine speed and accuracy is still a problem for this endeavor. A video-based object counting approach is used in this research. With portable cameras, videos of highway traffic are initially recorded. "Adaptive Transformer-based You Only Look Once (YOLO)-V5 (AT-YOLO-V5)" is then used to determine the sources, such as pedestrians and vehicles, from the recorded videos. Here, the "enhanced sailing fish optimizer algorithm" is used to optimize the parameters of the AT-YOLO-V5 network to increase the performance of the object counting task. The proposed traffic vehicle counting model's object detection and counting capabilities make it a perfect tool for tracking and evaluating data from a variety of sources, including "time, population, land use, traffic density, pedestrian flow, speed, and road dimensions”. This study assessed two machine learning techniques for forecasting traffic planning models and compared their accuracy through the R-coefficient and root mean square error. Gaussian processes and multilayer perceptrons were employed in experiments with real data, revealing that the multilayer perceptron demonstrated the highest average accuracy of 0.95% across all forecasting horizons.