Aero-engine blade defect detection method based on convolutional neural network

Abstract

The invention discloses an aero-engine blade defect detection method based on a convolutional neural network, and the method comprises the steps: changing a bilinear interpolation method in an up-sampling structure of a Yolov3 model into a bicubic interpolation method, improving the fusion effect of a bottom-layer feature and a deep-layer feature, and improving the detail perception capability ofa detection model; and meanwhile, adding an attention mechanism into a backbone network of the Yolov3 model, so that the feature expression capability of the region of interest is enhanced, the influence of useless information such as background noise on a detection result is reduced, and the region focusing capability of the detection model is improved. By means of the two methods, the improved Yolov3 model is obtained, the defects of the aero-engine blade can be detected through the improved Yolov3 model, the recall rate of a defect region detection frame in an image is effectively increased, and the detection precision of a defect region is improved.

Description

technical field [0001] The invention belongs to the technical field of machine vision, and in particular relates to a defect detection method. Background technique [0002] Surface defect detection of aircraft engine blades is an important task in aviation quality inspection and maintenance. Common defects include defects, craters, bending and corrosion, etc. Timely and effective detection of damage information can effectively avoid disasters caused by mechanical device damage. At present, defect inspection mainly relies on the human eye observation of ground crew, which is a tedious and labor-intensive work, and the work quality is easily affected by subjective factors. Therefore, in order to improve the detection efficiency and reduce the detection cost, an automatic defect detection method is needed. Traditional defect detection methods can be roughly divided into edge detection-based methods and morphology-based methods. In an ideal situation, traditional methods can d...

AI Technical Summary

Problems solved by technology

However, these methods will be affected by noise such as scratches on the actual blade surface and tiny pores in practical applications. The threshold segmentation method can be used to separate defective pixels from the global background. very sensitive to changes in light

To sum up, traditional defect detection algorithms are vulnerable and have limited room for improvement. It is necessary to develop a detection method for aero-engine blade defects in combination with neural network algorithms.

Images