Visual sensor based abnormal event detection with moving shadow removal in home healthcare applications

Area of Science:

• Computer Vision
• Artificial Intelligence
• Biomedical Engineering

Background:

• Vision-based abnormal event detection is crucial for home healthcare systems.
• Moving cast shadows often lead to misclassification in object detection and tracking.
• Accurate shadow removal is essential for reliable video surveillance and activity recognition.

Purpose of the Study:

• To design novel computer vision techniques for accurate object extraction and abnormal activity discrimination.
• To improve the accuracy of object detection and tracking through an effective shadow removal algorithm.
• To enhance abnormal event detection, specifically fall detection, using shape features and 3D trajectory analysis.

Main Methods:

• Development of a novel shadow removal algorithm to improve object detection accuracy.
• Utilizing shape features variation and 3D trajectory analysis for abnormal event detection.
• Implementation of computer vision techniques for detecting, tracking, and recognizing objects in home healthcare environments.

Main Results:

• The proposed shadow removal algorithm enhances object detection and tracking accuracy.
• Abnormal event detection achieved a 97% success rate with a 2% false positive rate.
• The system successfully distinguished various fall activities (forward, backward, sideways) from normal activities.

Conclusions:

• The developed computer vision techniques significantly improve abnormal event detection in home healthcare.
• The novel shadow removal algorithm and 3D trajectory analysis are effective in identifying falls and other critical events.
• This approach offers a reliable solution for enhancing safety and monitoring in home healthcare settings.