Identification method of motion blur code points based on convolutional neural network

Abstract

The invention discloses an identification method for a motion blurred coding point based on a convolutional neural network. The method comprises specific steps: 1, an actual photographing camera is calibrated, virtual camera coding points are built, and a large amount of motion blurred coding point image-coding point identity samples are acquired; 2, a convolutional neural network MBCNet is built; 3, the motion blurred coding point image-coding point identity sample set is used for training and testing the convolutional neural network MBCNet; and 4, the convolutional neural network MBCNet after training is used for carrying out segmentation and classification on the motion blurred image, and a corresponding coding mark point identity ID can be acquired. The identification method can process the actually-photographed motion blurred image, the corresponding coding mark point identity can be obtained, quick and reliable prophase data can be provided for machine vision measurement on a high-speed motion object, and the application field of the machine vision measurement method is expanded.

Description

technical field [0001] The invention belongs to the field of machine vision measurement, in particular to an identification method of motion fuzzy coding points based on a convolutional neural network. Background technique [0002] Coded markers are widely used in machine vision-based industrial metrology and reverse engineering. The identification of coded markers is the basis for subsequent processing. Existing coded marker identification methods are all based on clear images captured still. The recognition method generally includes the following steps: 1) Image noise reduction, 2) Edge detection, 3) Select a closed curve close to an ellipse from the edge as a candidate for the center of the encoding point, 4) Perform geometric transformation on the local image around the ellipse, It can be transformed into a circle or a rectangle. 5) In the transformed image, it is judged as 0 or 1 according to the gray level of the pixel. 6) It is decoded according to the binary string...

AI Technical Summary

Problems solved by technology

However, the result of increasing the sensitivity is that the signal-to-noise ratio is reduced, and serious noise will lead to a reduction in image quality, which is not conducive to post-processing

The result of increasing the aperture is that the depth of field becomes shallower, and the out-of-focus blur becomes more serious. Moreover, the aperture range of the lens is limited by the physical structure and manufacturing cost, which is generally very limited.

[0006] Therefore, the two methods of reducing the exposure time and increasing the aperture cannot solve the motion blur problem of the image when the measured object is in a high-speed motion state.

Images