Wearable Camera Video Analysis with Automatic Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for video analysis in wearable cameras are inefficient, requiring manual frame-by-frame review for mosaic processing, leading to increased workload and potential errors in identifying and masking individuals, which can compromise video admissibility as evidence.
Innovation Solution
A monitoring video analysis system that includes a camera and a server capable of detecting and displaying objects in each frame, allowing for automatic object detection and flexible masking, reducing the workload while maintaining privacy protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual frame-by-frame review is performed for mosaic processing, then privacy protection is ensured, but workload increases and processing time is extended
Solution Approach 1:
The system performs preliminary automatic object detection and masks candidate areas before final review. By pre-identifying potential subjects requiring privacy protection through AI detection, the system reduces the scope of manual verification needed, thereby protecting privacy while reducing processing time
Solution Approach 2:
The system introduces an automatic detection algorithm as an intermediary between raw video and final masked output. This intermediary layer pre-processes the video by identifying candidate subjects, allowing manual reviewers to focus only on verifying and adjusting AI-detected areas rather than examining every frame from scratch
2Productivity
If automatic mosaic processing is performed using dedicated image processing software, then processing efficiency is improved, but erroneous recognition occurs and admissibility as evidence decreases
Solution Approach 1:
The system implements feedback mechanisms where detected objects are displayed with bounding boxes and confidence scores, allowing users to verify and correct detections. The system also provides feedback on detection accuracy and allows iterative refinement, ensuring that automatic processing maintains high reliability for evidentiary purposes
Solution Approach 2:
The system performs partial automatic processing by detecting only high-confidence subjects or applying different processing levels to different frames. Rather than fully automating all aspects, it combines automatic detection with selective manual verification, achieving efficiency while maintaining admissibility through controlled automation
3Ease of operation
If mask disable area and movement pattern are preset, then processing is simplified, but flexibility to handle various video scenarios is reduced
Solution Approach 1:
The system uses dynamic object detection that adapts to different video scenarios in real-time. Rather than relying on fixed preset areas and movement patterns, the detection algorithm dynamically identifies subjects based on their actual appearance and behavior in each video, maintaining simplicity while achieving high adaptability across diverse situations
Data Source
AI summary
A monitoring video analysis system includes a wearable camera and a back end server that receives video data files of a captured video. The back end server detects an object appearing in frames constituting the received video data file, for each frame. The back end server stores a position of the object detected for each frame, for each object, as tracking information. The back end server displays the object detected for each frame, on monitor, by using identifiable solid-line border.