A New End-to-End Secondary Network for High-Efficient Vehicles and License Plates Detection

Up to now, a number of existing methods have worked fairly well in the field of vehicle and license plate detection. However, those methods are still not sufficient enough for satisfying practical industrial demands in terms of efficiency and accuracy, exhibiting poor performances such as incomplete detection of license plated when multiple vehicles are contained in the target image, and poor vehicle recognition when illumination is inadequate at night. To solve these limitations, a novel end-to-end secondary detection network named ES-Yolov3-tiny is proposed in this paper. ES-Yolov3-tiny can improve detection efficiency and accuracy by virtue of the secondary detection technology, and detect multiple vehicles in the same target image with the help of the interested region extraction technology. Compared with other existing state-of-the-art methods, the experiments show that the proposed method has better accuracy and lower computational cost.