DNN Classification Boundary Scope for AOI Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional classification methods for to-be-classified samples require extensive manual labor for labeled data generation, which is inefficient and time-consuming, especially in automated optical inspection (AOI) systems where false alarm rates are high.
Innovation Solution
A classification method utilizing a Deep Neural Network (DNN) to automatically extract feature responses from to-be-classified samples and determine their category based on a boundary scope of training samples, reducing the need for manual classification by integrating a determination unit and boundary learning unit to refine classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual classification is performed to obtain labeled data for defect classification, then classification accuracy can be improved, but labor hours and time consumption increase significantly
Solution Approach 1:
The system enables self-service classification by training the DNN model to automatically classify defect types without requiring manual classification for every sample. The model learns from initially labeled data and then performs autonomous classification, eliminating the need for continuous manual intervention while maintaining high accuracy
Solution Approach 2:
Labeled training data is prepared in advance to train the DNN model before deployment. This preliminary classification work is performed once during the training phase, enabling the model to perform rapid automatic classification thereafter without requiring ongoing manual labor
2Productivity
If DNN is used for automatic feature extraction and classification, then manual labor is reduced, but classification accuracy may deteriorate without proper boundary determination
Solution Approach 1:
A boundary determination unit is introduced as an intermediary between the DNN's automatic classification and the final classification result. This unit calculates boundary values based on feature responses and compares them against classification thresholds, serving as a mediator that enhances the accuracy of automatic classification without requiring manual intervention
Solution Approach 2:
The system implements feedback by using the boundary determination results to refine and adjust the classification decisions. The boundary values provide feedback information that validates or corrects the DNN's automatic classification output, ensuring high accuracy while maintaining automated operation
Data Source
AI summary
A classification method includes the following steps. Firstly, a classification module including a deep neural network (DNN) is provided. Then, to-be-classified sample is obtained. Then, the DNN automatically extracts a feature response of the to-be-classified sample. Then, whether the feature response of the to-be-classified sample falls within a boundary scope of several training samples is determined; wherein the training samples are classified into several categories. Then, if the feature response of the to-be-classified sample falls within the boundary scope, the DNN determines that to-be-classified sample belongs to which one of the categories according to the training samples.