Deep Learning-based accident image detection system
Deep-learning based accident image detection system detects moving & target objects within
CCTV surveillance range and automatically recognize and alert emergency situation such as
accidents, reverse driving and fire specified in the management guideline. Compared to computer
vision-based video analysis techniques, this AI based system is less influenced by weather or place and detects accidents with higher accuracy.
1) Technology Introduction : Deep Learning-based accident image detection system
Object Detection is a research field that analyzes video information by computer instead of a
person. Object detection technology is used variously in our daily life including pedestrian
detection, vehicle and road sign recognition by analyzing CCTV video. Especially, in the field of
traffic flow monitoring, the nationwide road network using CCTV is constantly monitored. In some
major section, an accident recognition system is in operation to automatically detect and inform
unexpected situations such as traffic jams, collision and falling objects and so forth. Typical
accident detection systems are equipped with algorithms based on traditional vision technology,
but it is highly sensitive to changes in surrounding environment such as luminance and weather,
resulting in pre-detection and detection errors. To solve this problem, DBNtech developed an
accident detection system that can detect objects in video like humans in various environments
using Deep Learning-based object detection technology. In collaboration with KICT(KOREA
INSTITUTE of CIVIL ENGINEERING and BUILDING TECHNOLOGY), we have completed an accident
detection system optimized for various environments and have set demonstrate-installed and delivered the system on since 2019.
2) Related images
3) Application of Deep Learning-based Object detection
A. Detection of unusual event (accident) through CCTV image analysis
1. Detection of stop(or collision) and reverse driving
2. Pedestrian detection
3. Fire and smoke detection
B. Character detection in images
1. Card number recognition
2. Vehicle’s license plate recognition
C. Object recognition in space
1. Product recognition for unattended store payment system
2. Automatic inventory check-up in refrigerator
D. Behavioral pattern analysis
1. Human motion tracking
2. Calculation of human’s action radius