Structured Light Paint Defect Detection With Mask-RCNN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional paint defect detection in vehicles is labor-intensive, subjective, and inconsistent due to expert evaluation, and existing automated systems are inaccurate and require large sample sets for training, making them inefficient for industrial use.
Innovation Solution
A deep learning-based system using phase-shifted patterns and machine learning algorithms, specifically Mask-RCNN, to detect and classify paint defects on vehicle surfaces by projecting multiple patterns and capturing images with a structured light source, enabling accurate and efficient defect identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If expert evaluation is used to detect paint defects, then detection accuracy can be maintained at a reasonable level, but the process becomes labor-intensive and time-consuming
Solution Approach 1:
The patent replaces the mechanical/manual inspection system with an automated optical and computational system. A projector displays patterns on the vehicle surface, cameras capture the pattern distortions caused by defects, and machine learning algorithms automatically analyze the captured images to detect paint defects, eliminating the need for manual expert evaluation while maintaining detection accuracy
Solution Approach 2:
The patent creates a digital copy of the inspection process by capturing images of the vehicle surface with projected patterns and storing them in a database. These digital copies are then processed by machine learning models to detect defects, replacing the physical act of expert visual inspection with automated digital analysis
2Productivity
If automated inspection systems are implemented, then inspection speed increases, but detection accuracy decreases due to subjectivity and bias in expert evaluation
Solution Approach 1:
The patent implements a feedback mechanism where captured images are fed into machine learning models that continuously learn from new data. The system compares detected defects against known defect patterns and refines its detection criteria over time, eliminating human subjectivity and ensuring consistent, objective defect detection across all inspections
Solution Approach 2:
The patent changes the detection parameters from human visual judgment to quantitative image analysis metrics. The system analyzes pattern distortion, image intensity, and spatial frequency characteristics to objectively determine defect presence, replacing subjective expert evaluation with measurable, reproducible parameters
3Measurement precision
If multiple patterns are projected to improve detection accuracy, then defect detection capability enhances, but the complexity of the inspection system increases
Solution Approach 1:
The patent makes the projected patterns serve multiple functions: they act as illumination sources to highlight surface defects, provide reference patterns for distortion analysis, and enable different detection modes by varying pattern characteristics. This multi-functionality reduces the need for separate inspection components and simplifies the overall system
Solution Approach 2:
The patent merges the illumination system with the inspection system by using the same projected patterns for both lighting the surface and providing detection references. The camera captures both the reflected light information and the pattern distortion data simultaneously, combining multiple measurement functions into a single integrated system
4Productivity
If existing automated systems are used, then inspection speed improves, but they require large sample sets for training and incur high costs
Solution Approach 1:
The patent applies partial action by using only the necessary pattern characteristics for defect detection rather than capturing all possible surface information. The system selectively analyzes specific frequency components and spatial regions where defects are most likely to appear, reducing the computational load and data processing requirements while maintaining detection effectiveness
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides superior accuracy in detecting and localizing paint defects with reduced data requirements and cost, achieving high-speed and consistent results in industrial settings.
Implementation Method 1
The projector is configured to display a plurality of patterns on a vehicle body at a plurality of different phases
Implementation Method 2
a camera configured to capture an image of the vehicle body after the pattern is projected onto the vehicle body
Data Source
AI summary
A method for detecting paint defects on objects is provided. The method comprises: projecting a plurality of patterns on a surface of an object at a plurality of different pattern characteristics; capturing a plurality of images of the object based on projecting the plurality of patterns; inputting the plurality of images of the object into a machine learning model to determine whether the surface of the object includes one or more paint defects; and based on determining, an image, of the plurality of images, includes a paint defect, causing display of the image with the paint defect.