PCB Component Detection Using Size-Segmented AI Classifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to effectively detect and identify a wide range of components on a printed circuit board (PCB) due to their varying sizes, which poses challenges in memory and processing limitations, leading to inaccurate and inefficient object detection.
Innovation Solution
Employ multiple pre-trained artificially intelligent classifiers to detect components of different size ranges, using a top-down approach to sequentially identify and exclude larger components before detecting smaller ones, and utilize a sliding window technique to generate scaled sub-images for training and detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single AI classifier is used to detect all components on a PCB, then the device complexity is reduced, but the measurement precision deteriorates due to inability to handle varying component sizes effectively
Solution Approach 1:
The patent divides the detection task into multiple segments by creating separate AI classifiers for different size ranges of components. Each classifier is specialized for detecting components within a specific size range, which improves detection accuracy for each size category while maintaining manageable system complexity through modular architecture.
2Measurement precision
If multiple AI classifiers for different size ranges are employed, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The detection system is segmented into multiple specialized classifiers, each handling a specific size range of components. This segmentation allows each classifier to be optimized for its specific task, improving overall detection precision while keeping individual classifier complexity manageable.
Solution Approach 2:
The patent implements a dynamic selection mechanism where the system automatically determines which classifiers to activate based on the detected component sizes. This dynamic approach allows the system to adapt to different PCB configurations and component distributions, improving precision without consistently activating all classifiers.
3Productivity
If all components are detected simultaneously without size-based grouping, then the productivity is improved, but the measurement precision deteriorates due to overlapping detection ranges
Solution Approach 1:
The patent segments the detection process into size-based stages, where classifiers process different size ranges sequentially or in parallel groups. This segmentation prevents detection conflicts and overlapping errors, improving precision while maintaining productivity through efficient stage-based processing.
Solution Approach 2:
The system performs preliminary sorting or grouping of components by size before detection, allowing classifiers to process components in an optimized sequence. This preliminary organization prevents detection conflicts and improves both accuracy and efficiency by avoiding redundant processing.
4Measurement precision
If a top-down approach is used to detect larger components first and exclude them from smaller component detection, then the measurement precision improves by avoiding false positives, but the processing time increases
Solution Approach 1:
The patent implements preliminary action by having larger component detectors run first and mark their detected regions, which then guides smaller component detectors to avoid those areas. This preliminary processing prevents false positives for smaller components while optimizing the overall detection sequence to minimize total processing time.
Solution Approach 2:
The detection system dynamically adjusts the processing sequence based on component size, with larger components processed first and smaller components processed subsequently in excluded regions. This dynamic ordering optimizes both precision by preventing false positives and time efficiency by processing components in a logical sequence.
Data Source
Figure 1~2
Figure 3~5
Figure 4A~4C
AI summary
The disclosure relates to apparatuses and methods for detecting elements of an assembly, such as electrical components in a printed circuit board. First and second artificially intelligent classifiers are provided for detecting elements in a high-resolution image of the assembly, wherein the first artificially intelligent classifier is pre-trained to detect first elements and the second artificially intelligent classifier is pre-trained to detect second elements, each of the first elements having a size within a first size range, and each of the second elements having a size within a second size range, in which the first size range includes elements having a size that is greater than the size of elements included within the second size range. The second artificially intelligent classifier can be prevented from subsequently searching for elements within bounding boxes previously obtained by the first artificially intelligent classifier.