Abstract
Campus security incidents occur from time to time, which seriously affect the public security. In recent years, the rapid development of artificial intelligence has brought technical support for campus intelligent security. In order to quickly recognize and locate dangerous targets on campus, an improved YOLOv3-Tiny model is proposed for dangerous target detection. Since the biggest advantage of this model is that it can achieve higher precision with very fewer parameters than YOLOv3-Tiny, it is one of the Tinier-YOLO models. In this paper, the dangerous targets include dangerous objects and dangerous actions. The main contributions of this work include the following: firstly, the detection of dangerous objects and dangerous actions is integrated into one model, and the model can achieve higher accuracy with fewer parameters. Secondly, to solve the problem of insufficient YOLOv3-Tiny target detection, a jump-join repetitious learning (JRL) structure is proposed, combined with the spatial pyramid pooling (SPP), which serves as the new backbone network of YOLOv3-Tiny and can accelerate the speed of feature extraction while integrating features of different scales. Finally, the soft-NMS and DIoU-NMS algorithm are combined to effectively reduce the missing detection when two targets are too close. Experimental tests on self-made datasets of dangerous targets show that the average MAP value of the JRL-YOLO algorithm is 85.03%, which increases by 3.22 percent compared with YOLOv3-Tiny. On the VOC2007 dataset, the proposed method has a 9.29 percent increase in detection accuracy compared to that using YOLOv3-Tiny and a 2.38 percent increase compared to that employing YOLOv4-Tiny, respectively. These results all evidence the great improvement in detection accuracy brought by the proposed method. Moreover, when testing the dataset of dangerous targets, the model size of JRL-YOLO is 5.84 M, which is about one-fifth of the size of YOLOv3-Tiny (33.1 M) and one-third of the size of YOLOv4-Tiny (22.4 M), separately.
Highlights
In the past few years, there have been occasional high-profile and horrific acts of violence on campus [1]
In this paper, inspired by DIoU-nonmaximal suppression (NMS) [30], we introduce the center distance ratio (CDR) of two detection boxes on the basis of the Gaussian weighted soft-NMS. e formula of CDR is as follows: ρ2b, bgt R c2, (4)
When the intersection over union (IOU) value of the two boxes is relatively large and the distance between the centers of the two boxes is relatively large, too, the two boxes are objects that are close to each other. is reduces the possibility of two close objects being misjudged as the same object detection box. e following experimental section shows that the phenomenon of missing detection is reduced effectively, and the evaluation result evidences that the proposed algorithm is more practical than state-of-the-art approaches
Summary
In the past few years, there have been occasional high-profile and horrific acts of violence on campus [1]. A new YOLOV3-Tiny backbone network is proposed to design a more lightweight and efficient campus dangerous target detection model. To redefine the campus of the dangerous objects and dangerous behaviors, the two types of targets are integrated into one model, which is lightweight and can achieve higher precision with fewer parameters. (3) Inspired by the DIoU-NMS algorithm, the soft-NMS algorithm is improved by adding the measurement of the center distance of the detection box, which has better performances than YOLOv3-Tiny’s own NMS It can effectively reduce the missing detection phenomenon when two targets are too close. Experimental results show that the jump-join repetitious learning YOLO (JRL-YOLO) has high cost performance in practical applications and can achieve higher precision with fewer parameters than YOLOv3-Tiny.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
