Defect detection method based on deep neural network heat map prediction

Abstract

The invention provides a defect detection method based on deep neural network heat map prediction. The defect detection method comprises the following steps: 1, preprocessing an acquired image; 2, predicting the heat map of the deep neural network; 3, performing defect point clustering analysis according to the thermodynamic diagram points output by prediction to obtain a defect set; 4, predictingthe defect type in the set; 5, aiming at different types of defects, extracting corresponding defect characteristics; and 6, carrying out grading processing on the defects according to the defect characteristics and the image. According to the method, the geometric features and the category information of the defects can be used for carrying out defect grading processing, and defect judgment is flexibly controlled according to the requirements of a production site. On the other hand, the function of the single model is dispersed in a plurality of modules, so that the method has better performance than the single model, and the detection speed of the whole process is more favorably optimized.

Description

technical field [0001] The invention relates to the technical field of defect detection, in particular to a defect detection method based on deep neural network heat map prediction. Background technique [0002] The surface defect detection of industrial products is an important link in the industrial production process, which is convenient for timely discovery of product and process defects and control of product quality. The surface defect detection technology based on machine vision has been widely used in all aspects of industrial production, which plays an important role in improving production efficiency and ensuring product quality in automated production. Surface defect detection based on machine vision includes two main steps. The first step is to perform high-definition imaging of the product to ensure that product defects can be displayed in the image. The second step is to analyze and process the image from the imaging device , to tell the user where there is a ...

AI Technical Summary

Problems solved by technology

The main reason is that, on the one hand, the technical solution is too simple, such as designing a single complex model to classify a fixed-size rectangular area in the input image, and judge whether the area has defects

On the other hand, the design of the technical solution is not deeply integrated with the actual situation of industrial production, which makes it difficult to implement the technical solution

The problems manifested are: the detection speed is slow, resulting in low production efficiency; the defect judgment scale cannot flexibly meet the changes in the yield rate of different customers and different production periods; the requirements for computing resources are high, usually 1 to 2 Nvidia 1080TI Only the GPU can meet the requirements of computing power; when switching products, a large amount of data needs to be collected and marked, which affects the efficiency of production

Images