Deep learning in vehicle detection using ResUNet-a Architecture

Abstract

Vehicle detection is still a challenge in object detection. Although there are many related research achievements, there is certainly room for improvement. This paper presents a deep convolution neural network architecture to identify vehicles. Detection performance can be improved by using the ResUNet-a architecture to extract features. Also, the use of edge detection using features that are detected from the deep convectional neural network. This work leads to a reduction in the number of calculations, and on the other hand. The removal of shadows, which leads to increased detection accuracy, can be one of the strengths of the method has been. Remove shadows are a critical step in improving vehicle detection. The method of combining color and contour features was used to remove shadows. Using the proposed method, the correct detection rate during the tests has improved the average accuracy, which shows that the proposed method in vehicle detection has had satisfactory and acceptable results. The results show that the accuracy and F-measure parameters are improved in this method.