Image surface defect detection method based on MobileNets

Abstract

The invention belongs to an image surface defect detection method, particularly relates to an image surface defect detection method based on MobileNets, solves the problems in the background technology, and adopts the technical scheme that comprises the following steps: creating an image training set and a category label; constructing a convolutional neural network; putting the created image training set and the category label into a convolutional neural network for learning and training; and detecting and classifying defects of the test pictures. The method is insensitive to image noise, andselection of the threshold has little influence on the segmentation effect; the selection of the filter type and parameters has little influence on the detection result, and the filtered image does not lose details; the method does not depend on characteristics of artificial design, and compared with a traditional algorithm, the method has good transportability and is not influenced by experienceof a designer; the network design not only pays attention to reduction of the parameter scale, but also gives consideration to optimization delay, is high in defect detection speed, and is more suitable for real-time online detection in an industrial environment.

Description

technical field [0001] The invention relates to an image surface defect detection method, in particular to an image surface defect detection method based on MobileNets. Background technique [0002] With the continuous improvement of my country's economic level, the application of computer vision, as an important part of future production technology, plays an important role in promoting the transformation of economic development mode, promoting the optimization and upgrading of industrial structure, and driving the development of high-tech. The traditional human eye detection method is difficult to meet the requirements of production efficiency, detection quality and low cost, so advanced detection methods must be used to ensure product quality, production efficiency and cost requirements at the same time. Among the many detection methods, the computer vision method is superior in detection speed, efficiency, cost and high flexibility, and is one of the most popular research...

AI Technical Summary

Problems solved by technology

Its advantage is that it has a wide range of applications, but its disadvantage is that the selection of filter type and parameters has a great influence on the detection results, and the filtered image is easy to lose details.

Traditional detection algorithms rely on artificially designed features, which always vary greatly depending on the detected object, so it is difficult to have good portability and is often limited by the designer's experience

Although the functions have become more and more complex, the detection effect has not improved significantly, and because of the excellent feature extraction ability of the neural network in deep learning, it has become a trend in recent years to apply it to the field of image processing.

Xiaojun Wu and other scholars designed a six-layer convolutional neural network based on the principle of GoogleNet to classify images and locate defects. Although it has a high accuracy rate for simple defect detection, the defect detection speed needs to be improved.

The FCN (Fully Converlution Network) network designed by Bian, X. can quickly fine-tune the model according to the detection data set, and has good flexibility, but the deepening of the network layer makes the computational complexity higher and higher

[0004] Although the algorithm based on deep learning is getting higher and higher in defect detection accuracy, due to the increasing network depth, the calculation complexity is also getting higher and higher, which requires high computing power of the platform

On the other hand, on-line detection at the industrial level requires extremely high detection speed, and the current detection speed needs to be improved

Images