In the production or testing of electronic products, it is often necessary to record the data of instruments in order to analyze the working state and performance of products. If the instrument does not have a communication interface, an instrument data acquisition system based on image recognition is designed to improve the efficiency of data acquisition in this paper. The system mainly includes camera, computer and software written in MATLAB. The software first acquires the image of the instrument through the camera, then chooses the area containing data in the instrument, and finally identifies the character of the selected area through the image recognition algorithm based on neural network and stores it in the editable text format. Through the test, the accuracy rate of the character recognition is more than 98%.