Introduction

The National Disaster Prevention and Rescue Science and Technology Center connects CCTV from different government departments, and the format is not standardized. The large number of anomalies cannot be manually identified one by one.

Therefore, this system uses artificial intelligence (AI) machine learning to develop the analysis mode of the image abnormality identification module and continuously corrects the parameters of the image interpretation algorithm to improve the effect of automatic image interpretation. After completion, the automatic image is built. The analysis module and warning module are used for long-term automatic monitoring and provide more efficient disaster warning when the image is abnormal, helping to shorten the disaster judgment time and improve the efficiency of disaster response operations.

Project performance and benefits

    AI image recognition module recognition targets include:
  • (1)Invalid image
  • (2)The road is flooded
  • (3)High water level of the river
  • (4)Road obstruction
  • (5)Bridge damage
  • (6)Displacement of non-structural objects

Use different AI model training methods to train the above six identification targets.