[Purposes]In order to meet the practical travel needs of blind people in various aspects,a ma-chine vision based guide stick has been designed.[Methods]The overall design of this system is modu-lar,with the core using STM32F407VET6 microcontroller and edge intelligent computing K210 chip;its machine vision function is achieved by using neural networks and YOLOv2 algorithm to train and estab-lish models,and detecting real-time images captured by cameras.The system itself has a gyroscope atti-tude angle measurement function,which determines whether the system rotation angle has reached the set threshold to determine whether the user has any signs of falling;the system also has GPS positioning function,and in case of emergencies,the system will send AT commands through the serial port to con-trol the GSM module to send location information to the emergency contact person.[Findings]All func-tions of this system have been simulated and tested,and the accuracy of machine vision function detec-tion can reach over 99%;the attitude angle measurement error is maintained within 1°;GPS and GSM functions can locate and send short messages.[Conclusions]The above functions can meet the basic practical travel needs of blind people.In addition,with the establishment of image detection models,fu-ture models can be trained to detect more objects,thereby improving machine vision functions and re-sponding to more actual travel situations.
NETL
NSTL
万方数据
