Rod and wire surface defect online detection device and method

Abstract

The invention relates to a rod and wire surface defect online detection device and method, a multispectral camera and a red, green, blue and near-infrared light source combined image acquisition device are used to synchronously acquire a surface gray image and a depth image of a to-be-detected rod and wire, and the surface defect of the rod and wire is detected through a gray image and depth imagefusion method. The multispectral camera acquires the reflective light of red, green, blue and near-infrared light sources on the surface of the rod and wire, red (R), green (G), blue (B) and near-infrared light (NIR) channel images are separated from the acquired images, the NIR channel image is taken as the gray image, the R, G and B channel images are used to construct a three-dimensional surface through a three-dimensional reconstruction algorithm, and the gray image is obtained. The surface defect area of the rod and wire is detected through the pixel-level fusion of the gray image and the depth image, the gray image and the depth image in the defect area are inputted into a convolutional neural network to classify the defect, and the defect identification result is obtained.

Description

technical field [0001] The invention relates to a method and an implementation device for on-line detection of surface defects of rods and wires by using multispectral imaging technology, belonging to the field of optical measurement. Through the combination of multi-spectral camera and red, green, blue, and near-infrared light sources, the grayscale image and depth image of the surface of the rod and wire are obtained simultaneously, and the surface defects of the rod and wire are detected and identified online through the fusion method of grayscale image and depth image to improve defect detection. and recognition accuracy. Background technique [0002] Surface defects of rods and wires are an important factor affecting the quality of rods and wires. At present, traditional non-destructive testing methods such as ultrasonic, infrared, magnetic leakage, and eddy current are mostly used to detect surface defects of rods and wires. These methods can only be used in offline or...

AI Technical Summary

Problems solved by technology

[0002] Surface defects of rods and wires are an important factor affecting the quality of rods and wires. At present, traditional non-destructive testing methods such as ultrasonic, infrared, magnetic leakage, and eddy current are mostly used to detect surface defects of rods and wires. These methods can only be used in offline or low-speed situations, and Detected defect types are very limited

With the development of CCD camera and computer image processing technology, surface inspection technology based on machine vision has received more and more attention. The two-dimensional detection method, this two-dimensional detection method is difficult to achieve good results in the application of rods and wires, the reasons are as follows:

[0003] (1) The section of the rod and wire is a curved surface. If the two-dimensional image detection method is used, there will be shadows and uneven illumination in the collected images, which will increase the difficulty of subsequent image processing.

[0004] (2) The surface of the rod and wire is covered with a large amount of iron oxide scale. The appearance of the oxide scale has no obvious rules, and there is no obvious difference from the real defects such as rolling scars, indentations, and inclusions on the surface of the rod and wire rod. It is difficult to distinguish

Due to the large amount of data in 3D detection, there is a contradiction between detection resolution and detection speed, and it is difficult to achieve the resolution and speed equivalent to 2D detection. If the two methods are not synchronized, the complementary advantages of the two cannot be realized.

[0007] (2) Shadow and uneven illumination in two-dimensional images

[0008] (3) Fusion of grayscale information and depth information

Traditional two-dimensional detection and three-dimensional detection are realized by different cameras, and the images obtained by different cameras do not correspond to the same area, and image registration is required, which increases the difficulty of information fusion

Images