Finding every car: a traffic surveillance multi-scale vehicle object detection method
- PDF / 1,148,384 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 33 Downloads / 207 Views
Finding every car: a traffic surveillance multi-scale vehicle object detection method Qi-Chao Mao 1 & Hong-Mei Sun 1,2 & Ling-Qun Zuo 1 & Rui-Sheng Jia 1,2
# Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract According to the problem that the multi-scale vehicle objects in traffic surveillance video are difficult to detect and the overlapping objects are prone to missed detection, an improved vehicle object detection method based on YOLOv3 was proposed. In order to extract feature more efficiently, we first use the inverted residuals technique to improve the convolutional layer of YOLOv3. To solve the multi-scale vehicle object detection problem, three spatial pyramid pooling(SPP) modules are added before each YOLO layer to obtain multi-scale information. In order to cope with the overlapping of vehicles in traffic videos, soft non maximum suppression (Soft-NMS) is used to replace non maximum suppression (NMS), thereby reducing the missing of predicted boxes due to vehicle overlaps. Our experiment results in the Car dataset and the KITTI dataset confirm that the proposed method achieves good detection results for vehicle objects of various scales in various scenes. Our method can meet the needs of practical applications better. Keywords Traffic surveillance . Vehicle object detection . YOLOv3 . Convolutional neural networks (CNNs)
1 Introduction Vehicle objects detection in traffic surveillance is an important issue in Intelligent Transport System (ITS). By detecting the vehicle objects in the surveillance video, it is possible to carry out follow-up work such as license plate recognition, vehicle tracking, and traffic flow statistics. Therefore, the vehicle object detection method is the premise and basis of these followup research work, and has a high application value. As a specific application of object detection technology, many experts have gradually begun research on vehicle detection. The original vehicle detection method mainly generates prediction regions based on prior knowledge, which mainly includes information such as left-right symmetry [1], vehicle shadow, and vertical edges [2]. The vehicle object detection
* Hong-Mei Sun [email protected] * Rui-Sheng Jia [email protected] 1
College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China
2
Shandong Province Key Laboratory of Wisdom Mine Information Technology, Shandong University of Science and Technology, Qingdao 266590, China
method is divided into traditional vehicle object detection and deep learning-based vehicle object detection. Traditional vehicle object detection is mainly by artificially extracting features and then using a classifier to discriminate whether the area belongs to a vehicle. For example, Felzenszwalb et al. proposed a Deformable Part Model (DPM) for object detection including automobiles [3, 4]. Dalal et al. proposed the use of Histogram of Gradient (HOG) features, combined with Support Vector Machine (SVM) for detection [5]. Karaimer et al. [6] pr
Data Loading...
