Work Recognition Switching Between Object and Skeleton Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing work recognition technologies face challenges in accurately estimating actions due to blind spots and erroneous detections, leading to reduced recognition accuracy, especially when objects are not detected or misdetected.
Innovation Solution
A work recognition system that switches between two recognition processes based on detection conditions, using first detection information for hands and objects, and second detection information for worker skeletons, to improve accuracy by selecting the most reliable method for each situation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection is performed using sensor data and coordinate association, then work recognition can be achieved, but recognition accuracy deteriorates due to erroneous detections and blind spots
Solution Approach 1:
The patent implements dynamic switching between two recognition methods based on real-time detection conditions. When object detection fails or produces erroneous results, the system automatically switches to skeleton-based recognition, and vice versa. This dynamic adaptation resolves the contradiction by ensuring reliable work recognition regardless of detection quality.
Solution Approach 2:
The system changes the recognition approach parameter based on detection reliability. Instead of using a fixed recognition method, it adjusts the method selection based on whether object detection succeeds or fails, thereby maintaining high accuracy across varying detection conditions.
2Device complexity
If only skeleton information is used for work recognition, then processing is simpler, but recognition accuracy deteriorates due to inability to distinguish specific work actions
Solution Approach 1:
The system dynamically selects between skeleton-only recognition and object-inclusive recognition based on whether objects are successfully detected. When objects are detected, the system uses both skeleton and object information for accurate work recognition. When objects are not detected, it falls back to skeleton-only recognition, thus balancing complexity and accuracy adaptively.
Solution Approach 2:
The recognition system is designed to perform multiple functions: it can recognize work actions using only skeleton information when objects are absent, and it can recognize work actions using both skeleton and object information when objects are present. This multi-functionality allows the system to maintain accuracy across different working conditions without requiring separate systems.
3Measurement precision
If comprehensive object and hand detection is performed, then work recognition accuracy can be improved, but processing complexity and time increase
Solution Approach 1:
The system implements dynamic processing depth adjustment. When object detection succeeds, it performs comprehensive analysis using both object and skeleton information for accurate work recognition. When object detection fails, it immediately switches to skeleton-only recognition, avoiding wasted processing time on failed detections and reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The system performs preliminary object detection before committing to comprehensive work recognition processing. This preliminary check allows it to avoid unnecessary complex processing when objects cannot be detected, thereby reducing processing time while preserving the option for accurate recognition when objects are present.
Data Source
AI summary
The present disclosure provides a work recognition device acquiring a photographed image capturing work of a worker; and detecting, based on the photographed image, first detection information relating to an object of the work and at least one of a right hand or a left hand of the worker.