Disaster warning system using Convolutional Neural Network - Focused on intelligent CCTV

Abstract

In this paper, we propose an intelligent CCTV technology which is applied to a recent attracted attention real-time object detection technology in a disaster alarm system. Natural disasters are rapidly increasing due to climate change (global warming). Various disaster alarm systems have been developed and operated to solve this problem. In this paper, we detect object through Neuron Network algorithm and test the difference from existing SVM classifier. Experimental results show that the proposed algorithm overcomes the limitations of existing object detection techniques and achieves higher detection performance by about 15%.

Fig. 1. water play safety system(Korea Coast Guard)

Fig. 2. climber safety system (National Disaster Management Research Institute)

Fig. 3. KISA intelligent certification process

Fig. 4. Common Object Detection Algorithm Flowchart

Fig. 5. HOG Algorithm

Fig. 6. Proposal Algorithm Flowchart

Fig. 7. YOLO v2 Network

Fig. 8. Clustering box dimensions on VOC and COCO

Fig. 9. Sutherland Hodgman Polygon Clipping

Fig. 10. Test Level-1 results (KISA Event)

Fig. 11. First event detection cumulative graph

Fig. 12. GT detection rate (all test algorithm)

Fig. 13. Test Level-2 results (GT)

Fig. 14. IoU Comparison by Test Algorithm 1, 2

Table 1. KISA Intelligent Authentication Evaluation Method and Criteria (Intrusion)

Table 2. Darknet 19

Table 3. Detection frameworks on PASCAL VOC 2007

Table 4. Test Dataset

Table 5. Test environment

Table 6. Test Video

Table 7. Test Level-1 (KISA Event)

Table 8. Test Level-2 (GT Event)

Table 9. Test Level-3 (IoU)