Design Of Blind Obstacle Avoidance System Based On ARM Technology

According to data from the China Disabled Persons 'Federation,the number of visually disabled patients in China reached more than 17 million in 2018,and the number of new blind people each year reached 450,000.China is one of the countries with the most blind people in the world.As a disadvantaged group,blind people often encounter difficulties in their lives that normal people cannot imagine.The data shows that about 30% of the blind people do not go out basically,and the other 46% need to be accompanied by family and friends.The problem of blind travel urgently needs people's attention.The irregular construction and management of blind roads in the city have led to the ineffective use of blind roads.At present,blind people mainly rely on traditional blinding methods in the traffic environment: blinding sticks can help blind people to detect the surrounding road surface,but the scope is limited,and it is impossible to find overhangs.Dangerous objects;guide dogs have a long training period and are expensive,which undoubtedly increases the pressure on the blind.Therefore,the traditional way of blinding cannot effectively solve the problem of blind travel and ensure their safety.The purpose of this article is to conduct research on the intelligent blind guide system.The purpose is to enable blind people to understand the road information like normal people when walking on the sidewalk,and to ensure that they can travel conveniently and safely.In this paper,through the study of blind guidance technology at home and abroad,the characteristics of the road environment are analyzed and the blind obstacle avoidance system based on ARM technology is designed in accordance with the needs of the blind.The embedded-cloud server architecture is used to ensure the real-time and reliability of the system.The design of the system mainly includes the underlying hardware design of the front end,the construction of the embedded platform,and the detection and identification of traffic lights and traffic signs.The front end uses the I.MX6U-ALPHA development board as the hardware platform.Its microprocessor is Cortex-A7 with I.MX6 ULL as the core.The embedded platform uses the open source Linux operating system.The traffic lights and traffic signs use traditional image detection technology.And deep learning-based methods for recognition.The main contents of this article are as follows:First of all,the overall design of the system is proposed,and how to build a Linux system development platform is introduced in detail;the cross-compilation tool chain is installed on the PC,and U-Boot transplantation,Linux kernel transplantation,root file system transplantation and equipment are carried out in this development environment The operation of the driver provides a stable and reliable embedded platform environment for the design of the blind obstacle avoidance system.The circuit design of the system hardware module data acquisition,its peripheral equipment modules include: image acquisition module,ultrasonic ranging module and voice broadcast module,the embedded front end is connected to the cloud server through WIFI,and the collected data is passed under the TCP / IP protocol.Socket socket for information exchange.Secondly,the graphic detection and recognition algorithm is studied in detail.Due to the lack of data sets and obvious color shape features of traffic lights,RGB to HSV color space is used to extract the candidate regions,and the morphological processing is used to remove noise.According to the shape characteristics of the signal lights,Hough is used.It is converted to shape detection,and the detected area is excluded from the non-signal light area by using the signal board information.Finally,the statistical information of the color histogram is used to complete the identification of the signal light.The detection and recognition of traffic signs use a deep convolutional neural network multi-task learning model.Based on the Linux system and Tensorflow1.4 architecture and the work of Zhu et al.,The network structure is improved.The MSRCR image enhancement algorithm is used after the image input layer.Image enhancement processing.A multi-scale strategy was used to intercept the fixed-size image blocks in the training and test phases and send them to the network for detection.At the same time,the multi-task learning model was reduced from the original 8 layers to 7 layers.Finally,the feasibility and real-time of the system in this paper were verified by actual measurements Sex.The system designed in this paper can effectively identify obstacles within the range of 1 to 3 meters of the blind person.According to the test results,the system still has a good detection effect for sign recognition under special conditions such as unsatisfactory lighting and obstruction.While the accuracy rate reached 91%,the real-time performance of the system was improved.Meet the requirements of smart devices for blind people going out.At the same time,the system of this article has certain reference significance for unmanned driving technology.