Holding Manner Learning for Commodity Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object recognition technologies in store systems struggle to accurately recognize commodities due to variations in holding manner, position, and distance between the image sensor and the commodity, leading to incorrect recognition.
Innovation Solution
An information processing apparatus with a designating module, acquisition module, calculation module, and updating module that designates a target object, acquires and compares image features, and updates a list based on similarity degrees to improve recognition accuracy by guiding proper holding manners for image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If general object recognition technology is used to recognize commodities, then the system can automatically identify objects, but the recognition accuracy deteriorates when the holding manner, position, or distance varies
Solution Approach 1:
The system performs preliminary learning by capturing images of commodities in various holding manners and storing them as reference data. This preliminary action enables the system to later accurately recognize commodities even when holding manners vary, as the reference database includes multiple examples of proper and improper holding positions and orientations.
Solution Approach 2:
The system provides feedback by displaying learned holding manner information to users. When a commodity is captured, the system compares the captured image with stored reference images and provides guidance information about proper holding manners, allowing users to correct their positioning for better recognition accuracy in subsequent attempts.
2Adaptability or versatility
If the image capturing section captures images in various holding manners, then more comprehensive data is collected, but the recognition accuracy for specific holding manners deteriorates
Solution Approach 1:
The system segments the holding manner space into distinct categories by capturing and storing images for different holding positions, orientations, and distances. This segmentation allows the system to later distinguish between different holding manners and provide accurate recognition for each specific category, while still maintaining overall comprehensiveness through the accumulated reference database.
Data Source
AI summary
A holding-manner learning apparatus includes a storage part and a hardware processor that functions as a designating module, an acquisition module, a calculation module, a storing module, and an output module. The designating module designates a target object from storage storing therein a feature amount for recognizing an object. The acquisition module acquires an image by photographing an object. The calculation module calculates a similarity degree between the feature amount of the object in the image and the feature amount of the target object. The storing module stores, according to the similarity degree calculated, the image and similarity degree in a first table or a second table in an associated manner. The output module outputs an electronic file presenting the image and similarity degree in the first table to indicate a proper holding manner and/or the image and similarity degree in the second table to indicate an improper holding manner.