Moving Image Recognition for Multi-Surface Code Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Point-of-sales (POS) systems struggle to associate data codes and labels when they are arranged on different surfaces of the same object, leading to difficulties in performing price reductions or discounts, as existing techniques require both to be captured in the same image.
Innovation Solution
A moving image recognition apparatus and method that includes a moving image input unit, buffer unit, moving object detection unit, data code reading unit, label recognition unit, and association unit, which detect and decode data codes and recognize labels, even when they are on different surfaces, by associating the recognition results and outputting them together.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If both data code and label are captured in the same image, then recognition accuracy is improved, but the system cannot handle labels on different surfaces
Solution Approach 1:
The system transitions from two-dimensional static image analysis to three-dimensional spatial-temporal analysis by tracking object movement across multiple frames. This allows the system to associate data codes and labels that appear on different surfaces by observing their relative positions and movements in three-dimensional space over time.
Solution Approach 2:
The system performs preliminary detection of both data codes and labels separately in the moving image sequence, then performs association processing in a subsequent step. This two-stage approach allows flexible handling of cases where targets appear in different frames or positions, resolving the contradiction between requiring simultaneous capture and handling different surface positions.
2Adaptability or versatility
If label is attached on different surface from data code, then versatility is improved, but association difficulty increases
Solution Approach 1:
The system employs dynamic tracking of object movement to associate data codes and labels. By continuously monitoring the movement trajectory of the object carrying both elements, the system can correlate their positions across different frames even when they appear on different surfaces, reducing association difficulty while maintaining placement flexibility.
Solution Approach 2:
The system uses the object itself as an intermediary to associate the data code and label. By detecting the object's movement and using it as a reference frame, the system can link the data code detected in one position with the label detected in another position, facilitating association despite different surface placements.
3Device complexity
If static image capture is used, then system simplicity is maintained, but price reduction processing fails for different surface labels
Solution Approach 1:
The system captures images periodically at fixed intervals to create a moving image sequence. This periodic capture approach maintains relatively simple system architecture while enabling the detection and association of data codes and labels that appear at different times in the sequence, thereby achieving price reduction processing for labels on different surfaces.
Solution Approach 2:
The system continuously processes the moving image sequence to detect and associate data codes and labels. This continuous processing approach maintains system simplicity by using a unified processing framework while enabling price reduction processing for various label placements by leveraging the temporal continuity of the moving image data.
Data Source
AI summary
According to an embodiment, a moving image recognition apparatus includes a moving object detection unit, a data code reading unit, a label recognition unit, an association unit, and an output unit. The moving object detection unit detects moving objects from a moving image stored in a buffer unit and identifies each of the moving objects. The data code reading unit detects a data code from each frame of the moving image and decodes the detected data code. The label recognition unit detects and recognizes a label from each frame of the moving image. When the recognized label and the decoded data code exist on the same object, the association unit associates them. The output unit outputs together the decoding result of the data code and the recognition result of the label associated with the decoding result.