A hyperspectral combined with deep learning industrial detection system

Abstract

The invention relates to a hyperspectral combined with deep learning industrial detection system, including a spectrometer for taking points and a hyperspectral camera for imaging; the detection steps are: S1, using the spectrometer to take defect points and normal points on the sample to be tested, And obtain the vector data of defective points and normal points; S2, use the algorithm based on comparing the mean square error between classes and between classes and calculate the correlation coefficient to verify the feasibility of hyperspectral detection; S3, if it is feasible, use the method based on measurement The method of selecting the eigenvalue gap between the class and the class is used to select the channel of the hyperspectral image; S4, using the method of comparing the eigenvalue of each point with the representative vector eigenvalue to automatically perform an automatic process on the image formed by the selected channel. Marking, sending the marked image into the neural network for training; S5, using the trained network to detect the object under test. The invention has the characteristics of simplicity, high efficiency and the like.

Description

technical field [0001] The invention relates to the technical field of visual inspection, in particular to a hyperspectral combined with deep learning industrial inspection system. Background technique [0002] The earliest hyperspectral detection is to see whether the defect area or the normal area has an absorption peak in which band after hyperspectral imaging. When only one of the two has an absorption peak in a certain band, only this band is used for imaging. Generate images with high defect contrast and use simple algorithms for detection. [0003] Later, a series of classification algorithms appeared, which can reduce the dimensionality of the pixel values ​​of multiple bands of each point in the hyperspectral image, and then bring them into the classification algorithm for classification. [0004] In recent years, deep learning convolutional neural networks have also been used in hyperspectral defect detection, which can be divided into supervised learning and unsu...

AI Technical Summary

Problems solved by technology

[0010] The technical problem to be solved by the present invention is: due to the large number of hyperspectral image channels, there is no ideal visualization method to see in a short time whether the defects on the current object under test can be detected by hyperspectral; The time, cost, and data complexity of hyperspectral imaging testing are a large overhead if feasible; if it is not feasible, it will be a waste of resources; and in the case of determining that hyperspectral imaging is feasible, if the defect and normal area are not If the absorption peak with high contrast is used, the classification algorithm will be very complicated, and the program cannot be parallelized, resulting in low program execution efficiency; if you try to use other convolutional neural networks with high parallelism and strong discrimination ability to achieve fast detection, In the training phase, due to the small amount of labeled samples and the hyperspectral image has almost no way to obtain a global visualization effect, it is difficult to obtain enough labeled data for effective training; therefore, the present invention provides a combination of hyperspectral depth Learn industrial inspection system to solve the above problems

Images