MEMS sensor defect type identification method and system

Abstract

The invention discloses an MEMS sensor defect type identification method and system. The method comprises the following steps: acquiring an MEMS sensor image; constructing a structure of combining an improved adversarial generative neural network with a convolutional neural network; training parameters of the improved adversarial neural network combined convolutional neural network structure; and inputting an MEMS sensor image to be detected into the trained network structure, and judging whether the input MEMS sensor image has defects or not. A plurality of groups of discriminators and generators are introduced into the improved generative adversarial neural network, one group of discriminator and generator is used for generating a defect picture, so that the defect picture is labeled after being generated by the generator and can be directly put into a subsequent convolutional neural network for training, the situation that an MEMS defect data set is small is exactly compensated, and meanwhile, classification of various different MEMS defect categories can be processed.

Description

technical field [0001] The application belongs to the technical field of image data processing, and specifically relates to a MEMS sensor surface defect detection method and system based on an improved adversarial generative neural network combined with a convolutional neural network. Background technique [0002] MEMS sensor is a micro-electromechanical system, which is a multi-disciplinary frontier research field developed on the basis of microelectronics technology. After more than 40 years of development, it has become one of the major scientific and technological fields attracting worldwide attention. It involves various disciplines and technologies such as electronics, machinery, materials, physics, chemistry, biology, and medicine, and has broad application prospects. As of 2010, there are more than 600 units in the world engaged in the development and production of MEMS, and have developed hundreds of products including miniature pressure sensors, acceleration senso...

AI Technical Summary

Problems solved by technology

It is precisely because of its small size that the error rate cannot be controlled by manual detection of defects on the upper surface of its manufacture. In addition, it plays an extremely important role in the product. If there is a problem with it, it will cause the function of the entire product to fail. Affected

If the convolutional network is used for identification alone, since the number of defect maps of the MEMS sensor is limited, the training set is not enough, which will also affect the recognition accuracy of the entire network.

Images