Digital micromirror highlight defect detection method based on deep learning

Abstract

The invention discloses a digital micromirror highlight defect detection method based on deep learning, belongs to the technical field of visual detection, and solves the problem of low highlight workpiece surface defect recognition efficiency. The method is established on a digital micromirror device calculation imaging system, strong reflection light on the surface of a highlight defect is modulated by the light intensity of a spatial light modulator, namely a digital micromirror device, and then enters a camera CMOS (Complementary Metal Oxide Semiconductor) surface to be imaged, and the light intensity modulation of the digital micromirror device uses a self-adaptive light intensity modulation method. A high dynamic range image data set is made by collecting a plurality of images, a deep learning target detection framework Mask RCNN is combined, a GPU is used for training a neural network, quick detection and segmentation of the highlight defect are finally achieved, the defect thata digital micro-mirror device or a deep learning network is independently used is overcome, and the detection accuracy of the highlight defect is improved. The detection efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of visual inspection, in particular to a detection method for highlighting defects of a digital micromirror based on deep learning. [0002] technical background [0003] Traditional computational imaging systems based on digital micromirror devices can achieve high dynamic range imaging and improve imaging clarity through reflective spatial light modulators, but the system does not have the ability to detect and identify defect pixel coordinates and defect types, and deep learning algorithms As an important research direction in the field of computer vision, it is difficult to classify and segment bright defect images purely in software. A high dynamic range image data set is produced through a computational imaging system based on a digital micromirror device. Combining the two, it can Realize the intelligent positioning, classification and segmentation of surface defects of highlighted workpieces, which ca...

AI Technical Summary

Problems solved by technology

[0003] Traditional computational imaging systems based on digital micromirror devices can achieve high dynamic range imaging and improve imaging clarity through reflective spatial light modulators, but the system does not have the ability to detect and identify defect pixel coordinates and defect types, and deep learning algorithms As an important research direction in the field of computer vision, it is difficult to classify and segment bright defect images purely in software. A high dynamic range image data set is produced through a computational imaging system based on a digital micromirror device. Combining the two, it can Realize the intelligent positioning, classification and segmentation of surface defects of highlighted workpieces, which can be applied to industrial fields where ordinary low dynamic cameras cannot detect and identify defects normally

Images