ISBN: 978-981-18-5852-9 DOI: 10.18178/wcse.2022.04.161
Real-time Detection System of Illegal Behaviors in Urban Management Based on Deep Learning
Abstract— The paper presents a real-time detection system of illegal behaviors in urban management based
on deep learning. First of all, we collected and annotated a large number of video images, and established
object detection dataset and ground image classification dataset, which serve as the data basis for the training
of the system's deep learning algorithm. Then, the algorithm framework of the system is proposed, and some
new methods are proposed in the key steps of the algorithm, such as upsampling method, filtering rules of
object tracking, ground ROI region extraction and illegal behavior determination. Experimental results
indicate that the proposed system can meet the requirements of practical applications in real time and
performance.
Index Terms— Machine learning, Computer vision, Deep learning, Object detection, Object tracking, Image
classification, Dataset.
Yanjun Fan
Research and Development Center, Suzhou Vortex information Technology Co., LTD., China
Zhuo Cheng
Research and Development Center, Suzhou Vortex information Technology Co., LTD., China
Chao Wang
Research and Development Center, Suzhou Vortex information Technology Co., LTD., China
Peng Tian
Research and Development Center, Suzhou Vortex information Technology Co., LTD., China
Cite: Yanjun Fan, Zhuo Cheng, Chao Wang, Peng Tian, " Real-time Detection System of Illegal Behaviors in Urban Management Based on Deep Learning, " WCSE 2022 Spring Event: 2022 9th International Conference on Industrial Engineering and Applications, pp. 1399-1407, Sanya, China, April 15-18, 2022.