LED bonding pad bubble AI detection method based on X-ray image

Abstract

The invention provides an LED bonding pad bubble AI detection method based on an X-ray image, and the method comprises the following steps: S1, building a semantic segmentation neural network model, setting at least two output channels, and forming a Loss function through the cross entropy loss of each output channel; S2, mixing the X-Ray images of at least two LED pads, and labeling the pads and bubbles in the images to form a training set; S3, training the neural network model built in the step S1 by using the training set formed in the step S2; S4, verifying the neural network model trained in the step S3, putting the neural network model which is verified to be qualified into use, and returning the neural network model which is verified to be unqualified to the step S2. A neural network semantic segmentation mode is adopted, the bonding pad background and the bubbles are quickly and effectively distinguished, the accuracy of a detection result under the conditions that the bonding pad background is complex and the bubble gray scale is uneven is greatly improved, and portability is high.

Description

technical field [0001] The invention relates to the technical field of bubble detection, in particular to an AI detection method for LED pad bubbles based on X-ray images. Background technique [0002] The traditional LED pad bubble detection method basically uses the gray difference between the pad and the bubble to identify. This bubble detection method has certain universality when the background of the pad is single and the gray scale of the bubble is uniform. However, the LED There are many categories and rapid replacement. In the actual inspection, LEDs have different imaging shapes under X-Ray, the pad background is complex, and the gray scale of the bubbles is uneven. Therefore, the traditional LED pad bubble detection method can detect under this background The effect is greatly limited and the accuracy of detection results is low. Contents of the invention [0003] The invention discloses an AI detection method for bubbles in LED pads based on X-ray images, whic...

AI Technical Summary

Problems solved by technology

[0003] The invention discloses an AI detection method for bubbles in LED pads based on X-ray images, which solves the problem of low accuracy of the detection results when the background of the pads is complex and the gray scale of the bubbles is uneven in the traditional LED pad bubble detection method. Adopt neural network semantic segmentation method to quickly and effectively distinguish pad background and air bubbles, greatly improve the accuracy of detection results in the case of complex pad backgrounds and uneven gray levels of bubbles, and have strong portability

Images