Chip pin missing detection method based on semi-supervised deep learning

Abstract

The invention discloses a chip pin missing detection method based on semi-supervised deep learning. Including: preprocessing the chip image to obtain an enhanced chip grayscale image; performing edge detection, thresholding, and morphological filtering on the enhanced chip grayscale image to obtain chip outline images and chip packaging images; chip outline images, Connected domain detection is performed on the chip package image, and it is judged that the side of the chip has pins; the package mask image is generated according to the judgment result of the side of the chip with pins, and the image of the chip pins is obtained by point-to-point multiplication with the chip outline image; The missing pins are judged and marked on the pin image; the semantic segmentation deep convolutional neural network is trained according to the marking results; the pin missing detection is performed on the image of the chip to be tested by using the trained semantic segmentation deep convolutional neural network. The present invention can realize missing detection of chip pins in the scene of chip quality detection, and improve detection efficiency and precision.

Description

technical field [0001] The invention relates to the technical fields of computer vision and deep learning, in particular to a chip pin missing detection method based on semi-supervised deep learning. Background technique [0002] In today's society, the status of circuit chips is very important. The scope of application is getting wider and wider, and the number of uses has also increased sharply. After the chips are produced, they need to be packaged with cores, peripheral circuits, and pins. The number of chips produced If there are too many, there will inevitably be many defective defects, and the missing pins will have a great impact on chip production, and have a great impact on the quality of the product and the reputation of the manufacturer. Because of the huge quantity, the cost of manual inspection will be very high, and intelligent inspection is an important way to reduce costs. [0003] In general, chips with the same function will be derived into many models. ...

AI Technical Summary

Problems solved by technology

The general detection method is template matching, but this method requires too high accuracy of thresholding segmentation, and a slight difference in the parameters of the light source and camera may lead to the failure of the entire system

Such systems require frequent manual maintenance, regular calibration and parameter adjustment

Moreover, the template matching method is too targeted and can only detect a specific chip, and there are restrictions on the type, model, and placement angle of the chip.

For other types of chips, the detection accuracy is low

The chip has the characteristics of small size, large production volume, and a large number of chips to be tested. The current chip detection technology detects the chips one by one, which is easy to miss, and the detection process is cumbersome and the detection efficiency is low.

[0004] Therefore, the existing chip pin missing detection technology has the problems of excessive specificity, difficulty in maintenance and use, low detection accuracy, and low detection efficiency.

Images