Image-Based Electrical Connector Assembly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine-learning routines for detecting electrical connector assemblies require large amounts of training data and resources, making them inefficient for determining connection states in manufacturing environments.
Innovation Solution
A control system using a deep learning neural network and region-based convolutional neural networks to define bounding boxes around electrical connector components, determining positional relationships between edges, and transmitting notifications based on connection states without the need for defect detection training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If machine-learning routines are used to detect electrical connector assemblies, then detection capability is provided, but large amounts of training data and computing resources are required
Solution Approach 1:
The patent extracts only the essential geometric features (bounding box edges) needed for connection state determination, eliminating the need for comprehensive defect detection training data. By focusing solely on positional relationships between connector components rather than general defect classification, the system requires minimal training data while maintaining detection capability.
Solution Approach 2:
The system uses simple geometric representations (bounding boxes) as copies of the actual connector components, replacing complex image-based machine learning models. These bounding box representations capture the essential spatial information needed for connection state determination without requiring resource-intensive training on full images.
2Reliability
If traditional machine-learning routines are used, then connection state detection is performed, but training iterations are excessive and resource consumption is high
Solution Approach 1:
The patent changes the parameter space from complex image features to simple geometric parameters (bounding box coordinates and edge positions). This parameter transformation enables accurate connection state detection with minimal training iterations, as the system only needs to learn spatial relationships between simplified geometric representations rather than complex visual patterns.
Data Source
AI summary
A method includes defining, in an image, bounding boxes about an electrical connector assembly, identifying an edge of each of the bounding boxes, and determining one or more metrics based on the edge of each the bounding boxes, where the one or more metrics indicate a positional relationship between the edges of the bounding boxes. The method includes determining a state of the electrical connector assembly based on the one or more metrics and transmitting a notification based on the state.