Design And Realization Of Automotive Instrument Symbol Recognition System Based On Convolutional Neural Network

A Automotive dashboard can show its own information to the driver,is the information center of automotive.Through the observation of the automotive instrument can understand the current status information,and then guide the driver to a safe driving.With the development of automotive is becoming more and more rapid,the development of automotive instrument is also being carried out at high speed.The shape and display effect of automotive instrument are also constantly evolving.The test of the automotive instrument becomes more and more difficult.The traditional manual test is not only inefficient,slow,easy to error by human participation,but also huge workload.Using science and technology to replace human labor is the most effective way to solve this problem.Using Convolutional Neural Networks to Automatically Identify automotive instrument Symbols is Another Application in the Field of image identification.With the continuous development of deep learning,The performance of neural networks has far exceeded human identification ability in some image identification issues.This is an important technical measure worthy of further study for the test problem of the automobile instrument.The main purpose of the subject is to research and analyze the application of convolution neural network in automotive instrument symbol recognition,and draw conclusions through experiments.The main tasks completed by the subject are:1.Describes the overall program of the symbol recognition system,the design and implementation methods of the convolutional neural network symbol recognition.The hardware environment of the identification system is established,and the optimization suggestions are put forward based on the advantages of the convolutional neural network method.2.Taking nearly a thousand dashboard pictures,and in accordance with the needs of the YOLO network model to mark the data for the identification of instrument symbols to prepare a reasonable and correct training data set and test data set.3.According to the dataset which is marked by man,YOLO network model is trained to identify the symbol of alarm lamp and the icon of display screen,respectively,and enough number of iterations are completed to achieve satisfactory recognition accuracy and recognition efficiency.4.Using the trained YOLO network model as the symbol recognition algorithm,the symbol features are extracted by convolutional neural network to realize the recognition of the indicator lamp of automotive instrument and the symbols of automotive instrument display icon,and show the position.5.Analyzing the advantage of image recognition based on convolution neural network algorithm compared with the traditional machine vision method and the improvement and optimization brought by convolutional neural network are studied under the circumstance of instrument symbol recognition.The final experiment and identification results are analyzed.The target detection algorithm based on convolutional neural network YOLO model can meet the actual system requirements,and can be applied and implemented in the automotive instrument symbol recognition system,and has good recognition effect and ideal real-time performance.For more instrument model expansion,recognition efficiency and future development,the convolution neural network based symbol recognition is superior to the traditional machine vision method.However,due to various reasons,there are still many deficiencies in the identification system.In the last chapter of this paper,the shortcomings and defects in the system are analyzed and summarized,and the next research direction is given.