Curved surface defect automatic detection method and apparatus thereof

Abstract

The invention discloses a curved surface defect automatic detection method and an apparatus thereof. The method comprises the following steps: 1, at a training phase, acquiring sample images, constructing a training set, performing manual defect identification on images in the training set, and marking areas where all defects appear; for each image in the training set, executing a defect pre-positioning step so as to obtain areas R where all the defects may appear; comparing the R with the manually marked areas where all the defects appear, and according to an overlap degree of the two, obtaining negative samples and positive samples through division; according to the positive samples and the negative samples, performing offline training of a deep neural network model, and outputting types of defect areas and concrete coordinates; and 2, at an online detection phase, acquiring current curved surface images to be detected, executing a defect pre-positioning step, obtaining sets R, and inputting the R into the network model, and obtaining the types and the concrete coordinates of the defect areas. The method and apparatus have the advantages of high adaptability, high real-time performance and high identification accuracy.

Description

technical field [0001] The invention relates to the field of image processing and deep learning research, in particular to an automatic detection method and device for curved surface defects. Background technique [0002] At present, the automatic detection of surface defects on small curved surfaces in the industrial field mainly uses the following algorithm: [0003] 1. Extract image features and perform image processing. This method generally uses industry experts to manually design image feature extraction methods according to defect characteristics, and then matches with real-time images; [0004] 2. Adopt the method of backpropagation neural network, use the artificially designed feature extraction method to extract image features and send them to the input end of the neural network, and classify whether the image has defects, defect types, etc. by establishing a neural network; [0005] 3. Using classifiers such as support vector machines to classify the manually ex...

AI Technical Summary

Problems solved by technology

[0007] 1. All the features are manually selected, which may cause the loss of useful information of the image due to inaccurate human subjective judgment, resulting in a decrease in recognition accuracy;

[0008] 2. The method of extracting features depends too much on the setting of parameters, and the applicability is not strong. For different types of defects, parameters often need to be reset;

[0009] 3. Classifiers such as neural networks can only classify the entire image, and cannot accurately locate the location of the defect, making it impossible to meet the requirements of online detection

Images