Research on the Model of Computer Vision Fusion Based on Microelectronics Technology

This paper proposes a quantum-inspired signal and feature fusion (QSFF) algorithm for deep fusion of microelectronic signals and computer vision features to improve target recognition performance. In the microelectronic module, the physical signal is preprocessed into digital data through operations such as sensor interface, amplification, filtering and analog-to-digital conversion; the computer vision module uses convolutional neural network (CNN) to extract image features. The QSFF algorithm nonlinearly fuses microelectronic signals and visual features through the superposition characteristics of quantum-like state vectors, and adaptively adjusts the complex-valued coefficients to enhance the information representation capability. The experiment uses a high-performance computing platform and a large complex scene target recognition dataset to compare the performance of the traditional fusion algorithm and the QSFF algorithm. The results show that the accuracy of the QSFF algorithm in complex scenes is significantly improved to 85%, which is 25% higher than that of the traditional algorithm; the recall rate on high-resolution images is stable at 80%, and the computing resource consumption and training time are greatly reduced, showing good computing efficiency and scalability. However, the model accuracy is slightly reduced in extreme lighting and occlusion scenes.