Neural Network Detergent Recognition for Automated Dispensing Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to accurately recognize detergent types from images of detergent containers due to variations in container arrangement, curvature, angle, distance, and lighting, leading to inefficiencies in detergent dispensing.
Innovation Solution
An electronic device uses a neural network trained on diverse images of detergent containers, including variations in angle, brightness, and blur, to recognize detergent information and guide the appropriate amount of detergent dispensed based on the recognized information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional character recognition methods are used to identify detergent types, then the system can guide detergent dispensing, but the recognition accuracy deteriorates due to variations in container arrangement, curvature, angle, distance, and lighting
Solution Approach 1:
The neural network is trained in advance using a large dataset of detergent container images captured under various conditions (different angles, distances, lighting, and container orientations). This preliminary training enables the system to recognize detergent types accurately despite variations in imaging conditions during actual use.
Solution Approach 2:
The system changes the approach from traditional character recognition to neural network-based image classification, transforming the recognition parameters from text-based to feature-based. The neural network learns to extract relevant features from images regardless of lighting, angle, or distance variations.
2Measurement precision
If a neural network trained on diverse images is used to recognize detergent information, then the recognition ratio improves, but the device complexity increases
Solution Approach 1:
The system uses image copying and data augmentation techniques to create diverse training datasets from limited sample images. By generating synthetic variations of detergent container images through copying and transforming existing images under different conditions, the neural network achieves high recognition accuracy without requiring extensive physical sample collection.
3Ease of operation
If character recognition through OCR is used, then the system can process text on detergent containers, but the recognition fails when characters are pictured or variously arranged
Solution Approach 1:
The system replaces the mechanical OCR text recognition process with a neural network-based image classification system. Instead of attempting to read and interpret text characters, the neural network directly classifies detergent types by learning visual patterns and features from images, making it robust to variations in text formatting, language, and character arrangements.
Data Source
AI summary
An electronic device includes a camera to capture an image, and a processor to input an image acquired by photographing a detergent container into a trained model to acquire detergent information corresponding to the detergent container, and to guide an amount of detergent dispensed based on washing information corresponding to the detergent information. The trained model is a neural network trained using images of a plurality of detergent containers.