Work Support Device Zone-Based Image Recognition for Worker Task Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing work management and instruction systems face challenges in accurately identifying worker operations due to individual differences in feature amounts and variations in workday behaviors, leading to reduced identification accuracy when using image recognition alone.
Innovation Solution
A work support device that specifies target work processes, selects relevant work areas, identifies worker tasks based on image data, determines deviations from work orders, and outputs determination results, thereby improving identification accuracy by focusing on specific work areas and tool usage patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image recognition is performed on the entire work table or all workers using a camera, then work identification can be attempted, but identification accuracy deteriorates due to individual differences in feature amounts and variations in workday behaviors
Solution Approach 1:
The patent divides the work table into multiple work areas and further divides each work area into multiple zones. Instead of performing image recognition on the entire work table, the system selectively images only specific zones where target works are located. This segmentation approach reduces the complexity of image recognition while improving identification accuracy by focusing computational resources on relevant areas only.
Solution Approach 2:
The patent applies different recognition strategies to different zones within work areas. High-precision image recognition is applied only to zones where critical works are performed, while other areas use simpler monitoring methods. This local quality approach ensures that computational complexity is concentrated where it provides the most value, thereby improving overall identification accuracy without uniformly increasing system complexity.
2Ease of manufacture
If simple image recognition is used to identify worker operations, then the system is easier to implement, but identification accuracy is reduced due to individual and daily variations in worker behavior
Solution Approach 1:
The patent pre-divides the work table into work areas and zones before work begins, and pre-identifies which zones require high-precision image recognition. This preliminary action allows the system to be implemented with simpler logic, as the recognition strategy is predetermined rather than dynamically decided during work execution. This approach maintains ease of implementation while improving accuracy by ensuring that critical areas are always monitored with appropriate precision.
Solution Approach 2:
The patent introduces work area division and zone identification as intermediary steps between simple image capture and complex operation recognition. These intermediaries structure the image recognition process, making it more manageable and accurate. The intermediary layer of spatial organization allows the system to maintain simplicity in implementation while achieving higher identification accuracy through structured, zone-based recognition.
Data Source
AI summary
The work support device stores information indicating a work order for each process, and information indicating a work area of a worker for each work in the process. A processor executes a specifying processing in which a target work to be performed by the worker is specified based on the information indicating the work order, a selection processing in which a target area where the target work specified by the specifying processing to be performed by the worker is selected based on the information indicating the work area, an identification processing in which a work performed by the worker is identified based on image data from a camera that captures an image of the area selected by the selection processing, a determination processing in which whether the worker has deviated from the work order is determined based on an identification result by the identification processing.