Research On Fast Instrument Recognition Algorithm Based On Embedded System

Pointer instrument is widely used in industry and daily life.At present,most of its reading methods are still manual reading,which is not only time-consuming,but also affected by the subjective judgment of the reader.Therefore,in order to improve the reliability and efficiency of instrument reading,this thesis studied the fast recognition algorithm of pointer instrument based on embedded system,and designed an embedded system for pointer instrument recognition.Aiming at the problem of reading recognition of pointer instrument,this thesis studied and analyzed the algorithm of pointer instrument recognition in complex environment,and determined the method of combining deep learning and image processing to quickly recognize the reading of pointer instrument.The main work of this thesis was as follows.(1)The embedded hardware platform of the pointer instrument rapid identification system was designed.The design of embedded hardware platform was based on the core board of Jetson nano.According to the requirements of pointer instrument recognition system,the hardware circuit design of system power supply,image acquisition,human-computer interaction,USB interface,TF card,Ethernet and Wi Fi was completed.(2)In the reading recognition algorithm of instrument,aiming at the problems of small target,poor detection effect and low reading accuracy caused by inclination of shooting angle of pointer instrument in complex background environment,a reading recognition method of pointer instrument based on deep learning and image processing was proposed.In this recognition method,YOLO V4 target detection algorithm was used to detect and extract the instrument panel in complex environment;after preprocessing,the instrument panel was corrected by image registration and perspective transformation;finally,the template matching algorithm was used to achieve the reading of the dial counting area,and Canny edge detection algorithm and Hough transform were used to determine the center of the circle and the pointer features to get the reading of the pointer.(3)The system software was designed based on the theory of instrument reading recognition algorithm.The system was modularized according to the functions needed by the pointer instrument rapid identification system,and the specific implementation process and program of each module were designed.Through the construction of the embedded system environment and the debugging of the system software and hardware,the embedded hardware platform could run stably,and the pointer fast identification algorithm designed in this thesis could effectively get the reading of the pointer.After the performance test,the average time of the embedded system for instrument identification was 1.563 s,which meets the requirements of rapidity to a certain extent.At the same time,the embedded system in the identification of 300 instrument pictures,the correct rate reached 97%,with good accuracy.The thesis has 64 figures,9 tables,and 57 references.