Research Report About Improving The Intelligent Drive By Image Stabilization and Radar Technology

Research Report About Improving The Intelligent Drive By Image Stabilization and Radar Technology

电子科技大学 格拉斯哥学院 张立澄 2017200602011

Background
The freshman seminar, in UoG-UESTC Joint School , invited some professors in the electronic engineering field to introduce the development of some electronic engineering technologies and many practical applications, with their professional knowledge and scientific research projects. I was impressed with visual processing, image stabilization, radar technology, etc. and inspired to utilize them to improve the intelligent transportation.

Introduction
Nowadays, the relatively low accuracy, stability and reliability of the radar and computer vision hinder the development of intelligent drive. It is helpful to solve this problem by making radar technology and computer vision technology be combined with professional visual processing and image stabilization technology. Through visual processing technology and the related algorithm to enhance the quality of the images, to improve the resolution of them and even achieve the image restoration of occlusion. While using the photo and video image stabilization technology by image superposition algorithm, eliminate jitter to improve the stability of the image.

Research Content
Since I participated in the robot competition to design wheeled robot based on SLAM technology. The successful achievement of simultaneous localization, map building and other functions enabled me to know the radar and SLAM technology. Besides, I joined an innovation and entrepreneurship team at our university, participating in a project about precise positioning of intelligent transportation, which makes me know something about the technical solutions (such as LIDAR, computer vision and GNSS) and industrial development of intelligent transportation. Most of the existing solutions are limited by the complex driving environment in real-world. For example, GNSS cannot achieve high-precision positioning under the shelter of buildings such as viaducts and tunnels, while computer vision cannot completely make sure high-precision and high-stability under the influence of variable driving conditions and weather environment.
With what I have learned from the freshman seminar, I consider that visual processing, image stability and radar technology can be combined to achieve high-precision and highly reliable positioning and identification in intelligent driving and even intelligent transportation. Using radar to achieve simultaneous localization and map building function, which are like cars’ “eyes”. At the same time, utilize the visual processing and image stabilization technology to strengthen the stability, reliability and precision of computer vision, which makes “bright eyes" to improve the development of intelligent transportation industry.
So, I did a huge variety of surveys, and found that there was already a similar solution for self-driving cars, launched by Ford Motor Company. The solution called "laser radar system”, was installed on the roof of cars, keeping launching the weak laser beam, to outline the simultaneous 3-dimension street view around the cars, and at the same time the cameras keep rotating 360 degrees to helps auto observe the surrounding environment. This system will analyze the collected information, and distinguish the distinction between constant objects (lane space, exit ramps, park benches, etc.) and the moving objects (the frightened deers, pedestrians and oncoming vehicles, etc.), and collect all the data together. And then based on the algorithm, designed by the university of Michigan, to judge the surrounding environment, and make corresponding responds.
However, the real-world driving and road conditions are not always constant and in good condition, for example, bumpy roads and bad weather are easy to affect the information collected by radar and camera. And there is no doubt that the situation of intelligent transportation would be greatly improved if radar technology and computer vision technology were combined with professional visual processing and image stabilization technology. That is to say, cars are equipped with radar and computer vision technology. The radar sends radar signals continuously to achieve simultaneous positioning and delineate the surrounding environment in real time. Computer vision utilizes the convolutional neural network to send the image into the network, and then the network classifies the image data. After that, object detection, target tracking (with generation algorithm and discrimination algorithm), and segmentation are carried out successively. What’s more, the computer vision technology with professional visual processing and image stabilization technology, which means through visual processing technology and the related algorithm to enhance the quality of the images, improve the resolution of them. Especially for some real-world situations, such as the overexposure or the dim light will lead to very low image contrast, and inaccurate identification, which is really dangerous for pilotless automobile in the real-world traffic. And it may be possible to use more advanced visual processing technology, combined with probability statistics and machine learning, etc., to achieve the image restoration of occlusion, which means inferring the content of damaged area in the imagine, according to the characteristic information left by the image, including color and structure. At the same time, using the image stabilization technology to make the image taken by camera more stable. Try to imagine, there must be some bumps during the real-world driving. What our dean taught in freshman seminar, the photo and video image stabilization technology, can be used to solve these problems exactly, according to a series of probability and statistics, such as image superposition algorithm, eliminate jitter. Improve the stability of the image is helpful to improve the high accuracy, stability and reliability of the computer vision identification, and it also make a huge contribution to the development of intelligent drive.

Conclusion
Combining radar and computer vision technology with professional visual processing and image stabilization technology can improve the high accuracy, stability and reliability of identification in intelligent drive. The computer vision technology with professional visual processing and image stabilization technology, which means through visual processing technology and the related algorithm to enhance the quality of the images, to improve the resolution of them and even achieve the image restoration of occlusion. Besides, using the photo and video image stabilization technology by image superposition algorithm, eliminate jitter to improve the stability of the image is helpful to improve the high accuracy, stability and reliability of the computer vision identification, and it also make a huge contribution to the development of intelligent drive.