研究目的
To find a possible solution for the visual inspection of welds by examining and testing two feature extraction methods in conjunction with two different classifiers to detect flaws and defects in a weld merely by inspecting the surface of the object.
研究成果
The proposed method achieves a high classification accuracy of 96% using an SVM classifier with features extracted by uniform LBP without rotation invariance, demonstrating the potential for automatic visual inspection of weld seams.
研究不足
The method's accuracy may be affected by the orientation of the weld in the images and the presence of unnecessary noise around the weld face. Further testing with different settings and classifiers could improve results.
1：Experimental Design and Method Selection:
The method involves segmentation, feature extraction, and classification using GLCM and LBP as feature descriptors followed by SVM and KNN classifiers.
2：Sample Selection and Data Sources:
A dataset of 100 weld seam images (50 good and 50 bad welds) was created for this purpose.
3：List of Experimental Equipment and Materials:
Nikon D3100 DSLR camera mounted with 18-55mm zoom lens was used for image capture.
4：Experimental Procedures and Operational Workflow:
Images were cropped, converted to grayscale, and processed for feature extraction before classification.
5：Data Analysis Methods:
Performance was evaluated using k-fold cross validation with k=5.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
