Abstract
Cylinder liner plays an important role in the internal combustion engine. The surface defects of cylinder liner will directly affect the safety and service life of the internal combustion engine. At present, the surface defect detection of cylinder liner mainly relies on manual visual inspection, which is easily affected by subjective factors of inspectors. Aiming at the bottleneck of traditional visual inspection technology in appearance inspection, this paper proposes a surface defect detection algorithm based on deep learning to realize defect location and classification. Based on the characteristics of the research object in this paper, the surface defect detection algorithm based on the improved YOLOv4 model is proposed, the model framework is constructed, and the data enhancement method and verification method are proposed. Experiments show that the proposed method can improve the detection accuracy and speed and can meet the requirements of the nonburr cylinder surface defect detection. At the same time, the method can be extended to other surface defect detection applications.
Highlights
Surface defects will directly affect the quality of the product, further affecting the chemical and physical properties of the product surface
As an important component of the internal combustion engine, the appearance of cylinder liner surface defects such as cracks and air holes will mean that there are major internal quality problems in the cylinder liner, which may lead to abnormal operation of the internal combustion engine and lead to safety problems. erefore, manufacturers and users put forward higher and higher requirements for the appearance quality of cylinder liner
The detection of cylinder liner surface quality mainly depends on manual detection. e manual detection method cannot meet the production needs in terms of work efficiency and is affected by the subjective experience of detection personnel
Summary
Surface defects will directly affect the quality of the product, further affecting the chemical and physical properties of the product surface. Different from R-CNN, Fast R-CNN inputs the whole image into CNN for calculation, and under the effect of ROI pooling, it fixes the output of CNN to a certain size of eigenvector In this model, classification and regression are implemented in different networks, so the detection accuracy is high, the detection speed is low. The two-stage target detection model has been improved a lot and the detection accuracy has been greatly improved, due to its complex model, the parameters of the model are too many and the training time is too long This kind of algorithm divides the classification and regression into two parts, which leads to low time efficiency in calculation. In order to improve the accuracy of the model, based on the YOLOv1, the YOLOv2 algorithm is improved by introducing Batch Normalization, anchor box mechanism, Input image
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
