Investigates Image data fusion techniques that combine image and track data from multiple sensors to achieve improved accuracies and more specific inferences than could be achieved by the use of a single sensor alone. Our aim is to explore the state-of-the-art image processing algorithms for achieving effective data fusion as in:
1. Introduction
Development of Abnormal Status/Behavior Determination Algorithm
The need for quarantine is emerging because immigration facilities that allow overseas inflow of infectious diseases and multi-weak facilities that allow secondary spread of infectious diseases are vulnerable to exposure to infectious diseases. One promising approach is to Develop an intelligent system that detects infection at all times and automatically classifies suspected infections safely. The goal is to establish an intelligent constant detection system that can suppress the spread and exposure of infectious diseases through detection and isolation of infected people and quickly, accurately, and simultaneously determine suspected symptoms of infection of individuals. Through this, it is possible to suppress the spread and exposure of infectious diseases through detection and isolation of infected people and to detect abnormal behavior of sus-objects.
2. Main Algorithm and Principle
- The proposed approach
The proposed model consists of an unsupervised learning-based detection model, supervised-anomaly image level behavior classification through object detection, and a skeleton-based behavior recognition research to integrated model.
- Spatio-temporal Action Detection
This task is to localize action instances and classify action categories in untrimmed video. Spatio-temporal action detection aims to localize action instances in both space and time, and recognize the action labels.
- unsupervised learning-based detection model
This task is to predict next frame based on input images and learned aspects, and several aspects of steady-state images are stored in memory modules and applied to the limited behavior patterns of people at airport/port immigration facilities.
- skeleton action recognition model and integrated model
This task is to utilize skeleton information to track people’s id and perform action recognition. In addition, propose a final behavior recognition model by presenting an integrated model using the above two studies.
3. Results
