Summary
In recent years, the rapid development of the automobile industry has led to a rapid increase in the number of automobiles in my country, but the number of casualties caused by traffic accidents remains high. From the perspective of protecting personal safety and reducing the possibility of traffic accidents, the vehicle anti-collision system can enable the driver to take automatic emergency braking measures when he does not notice the risk of collision with the vehicle in front, so as to avoid collision with the vehicle in front Or slow down the harm caused by the direct collision with the vehicle in front, which is more and more favored by the public. The key technology of the vehicle anti-collision system is to rely on the sensor installed on the front of the vehicle body to detect the target in front, accurately obtain the effective information of the target in front of the vehicle, and can track the target in front of the vehicle to predict its next movement state and then take corresponding measures . With the increasing requirement of automobile safety performance, more and more research institutions and automobile enterprises pay more and more attention to the research of vehicle anti-collision system.
In order to solve the problems of information sensing, target detection and recognition and tracking of the vehicle anti-collision system, this paper based on the sensor fusion algorithm, carried out the following work:
First of all, the difference between the sensors used for target detection in the vehicle anti-collision system is compared. In this paper, radar and vision fusion are selected as the sensors of the vehicle anti-collision system, and the main functions of millimeter-wave radar and vision in the vehicle anti-collision system are introduced in detail. Function.
And a method for accurate detection of targets in front of the vehicle based on millimeter-wave radar and visual camera is designed. Through the data preprocessing of the output targets of millimeter-wave radar and visual camera, it can accurately identify the target in front, and the test is carried out verify.
Secondly, according to the characteristics of the car, the motion state of the vehicle in front of the car is analyzed and studied, and the target motion model of the vehicle in front of the car is established. The characteristics of several target motion models are compared, and then the appropriate target motion model is selected for theoretical analysis of target tracking. According to the motion characteristics of the target in front of the vehicle, a simulation model is built in the Simulink software environment, and the target tracking simulation comparison test is carried out.
Finally, under the premise of ensuring the installation accuracy of the radar and camera, a real vehicle test platform is built for the vehicle anti-collision system based on the millimeter wave radar and visual camera, and the real vehicle tracking is performed on the target in front of the vehicle using the Kalman filter tracking algorithm based on the current statistical model Experimental results show that the target tracking algorithm used in this paper works well and can accurately track the vehicle ahead. The anti-collision test is carried out based on the target algorithm, and the real vehicle test is carried out with a stationary target as the front target, which verifies the effectiveness of the vehicle anti-collision system and its algorithm.
Key words: vehicle anti-collision; millimeter wave radar; visual camera; fusion algorithm; simulation
Abstract
In recent years, the rapid development of the automobile industry has made the total number of vehicles in our country increase rapidly, but the number of casualties caused by traffic accidents is still high. From the perspective of protecting personal safety and reducing the possibility of traffic accidents, vehicle anti-collision system can enable drivers to take automatic emergency braking measures when they are unaware of the danger of collision with the vehicle in front, so as to avoid collision with the vehicle in front or mitigate the harm caused by direct collision with the vehicle in front, which is increasingly favored by the public. The key technology of vehicle anti-collision system relies on its sensor installed in the front of the body to detect the target in front of the vehicle, accurately obtain the effective information of the target in front of the vehicle, and can track the target in front of the next motion state and then take corresponding measures. With the increasing demand of automobile safety performance, more and more research institutions and automobile enterprises pay attention to the research of vehicle anti-collision system.
In order to solve the problem of information sensing, target detection, identification and tracking of vehicle anti-collision system, this paper carried out the following work based on sensor fusion algorithm:
Firstly, the difference of sensors used for target detection in vehicle anti-collision system is compared. In this paper, radar and vision fusion are selected as the sensors of vehicle anti-collision system, and the main functions of millimeter-wave radar and vision in vehicle anti-collision system are introduced in detail.
A precise detection method of vehicle forward target based on millimeter-wave radar and vision camera is designed. By preprocessing the data of the target output by millimeter-wave radar and vision camera, it can identify the target in front of the vehicle accurately, and it is verified by experiments.
Secondly, according to the driving characteristics of the vehicle, the motion state of the vehicle in front of the vehicle is analyzed and studied, and the target motion model of the vehicle in front of the vehicle is established. The characteristics of several target motion models are compared, and then the appropriate target motion model is selected for the theoretical analysis of target tracking. According to the characteristics of vehicle forward target movement, a simulation model was built under Simulink software environment, and the target tracking simulation and comparison test were carried out.
Keywords: Vehicle collision prevention; Millimeter wave radar; Visual camera; Fusion algorithm; simulation
1.4 Introduction to simulation tool MATLAB/Simulink................................... .......... 4
4.3.2 Types of detect-then-track algorithms................................... .......... 19
4.5.1 3D model-based approach................................... ....................... twenty two
4.52 Method based on Kalman filter................................... ............. twenty three
Chapter 1 Introduction
1.1 Research topic background
my country is a big country with cars. Especially in recent years, the rapid economic development has led to a rapid increase in the number of cars in China. Cars have become the primary tool for people to travel. The popularity of automobiles not only makes people's daily life quick and convenient, but also brings a series of traffic problems. According to the latest research report jointly released by my country's State Administration of Work Safety and the Ministry of Transport, my country received a total of 8.643 million road traffic accidents in 2016, a year-on-year increase of 659,000, or 16.5%. Among them, there were more than 200,000 accidents that caused personal harm, resulting in more than 60,000 deaths, more than 200,000 injuries, and property losses as high as 1.21 billion yuan. The occurrence of rear-end collision accidents is one of the reasons for the huge loss of personal life and property safety. If the driver can be reminded to take measures or the car can automatically brake to avoid collisions before the rear-end collision accident occurs, most of the casualties will be avoided. will happen. Therefore, it is very important to research and develop a car collision avoidance system that can detect effective information in front of the vehicle, accurately track the target, and brake the vehicle when necessary, whether it is to protect the person from injury or to avoid traffic accidents. of.
1.2 Development status at home and abroad
Since the 1980s, foreign countries have made good achievements in the field of ADAS, mainly Google, Uber, Mobileye, Tesla and other companies as the main force to carry out related research. In particular, Israel's MobileEye has developed a complete set of advanced driver assistance system products. The forward collision warning developed by MobileEye is based on pure vision. After detecting the vehicle in front, it calculates the time to collision (Time To Collision, TTC). When the TTC reaches the set warning threshold, it will give a danger warning prompt. Warn up to 2.7 seconds in advance. At CES 2019, Mobieye announced that it will use its latest EyeQ5 chip in the intelligent driving system, focusing on product landing issues, and re-emphasized the importance of ADAS. At the same time, Mobieye also announced that it will cooperate with Great Wall Motors. In 2019, ADAS and higher-level intelligent driving systems were jointly developed according to China's traffic scenarios. The Autopilot automatic assisted driving system developed by Tesla uses a wealth of sensors - 8 cameras are installed around the body, which can achieve a 360-degree field of view without dead angles, and the detection distance can reach up to 250 meters; There are multiple ultrasonic sensors that complement the complete vision system. Nvidia is also investing a lot of energy in the direction of ADAS. In the 2019 CES, Nvidia released a clear signal that the technology for autonomous driving will be transferred to the Level2+ ADAS system. The core of Nvidia’s Level2+-based ADAS system is The self-developed high-performance Driver AGX Xavier has powerful computing power, can handle deep convolutional networks, and performs well in perception. For the use of sensors in ADAS, it is possible to choose monocular, binocular vision, millimeter wave radar or lidar according to different tasks. For example, Budzan S and others use 3D lidar to scan obstacles on the road, and ZiebinskiA et al believe that the fusion of multiple sensors is the future trend through research.
As early as the end of the 20th century, my country's vehicle automatic driving technology was still very backward. Due to the gradual realization of the backwardness and lack of precision, it gradually began to pay attention to the development of vehicle automatic driving technology. Around the beginning of the 21st century, a number of vehicle automatic driving systems began to emerge on the market, but the technology was quite backward, and only a very small number of low-level areas were in use. However, due to major innovations in the functions and performance of vehicle automatic driving systems at home and abroad, this has also laid a solid foundation for the innovation of vehicle automatic driving technology in my country. However, foreign countries are relatively advanced. In the 1980s, some European countries already had advanced vehicle automatic driving technology with high precision. By the mid-1990s, most of the old-fashioned automatic vehicle Advanced vehicle self-driving technology supersedes. Nowadays, the development direction of vehicle automatic driving technology is in the hands of our new generation of young people, and we must move forward rapidly in the direction of intelligent control, program control, and precision.
1.3 Significance of the subject research
Casualties and property losses caused by the increase in automobile accidents have become a social problem that cannot be ignored. Improving the safe driving guarantee capability of automobiles has also become the research and development focus of major automobile manufacturers. However, traditional passive safety technology not only cannot avoid traffic accidents, but also may cause secondary damage to personal safety, so the concept of active safety has gradually formed and is constantly being improved. Hyundai's active safety technology mainly includes electronic stability program (Electronic Stability Program, ESP for short) and advanced driver assistance system (Advanced Driver Assistance System, ADAS for short). To detect the surrounding information of the vehicle body, and then its control system judges whether the vehicle is in a safe environment by processing the information. If the processed information shows that the own vehicle is in danger, it can send a danger signal to the driver through its control system to assist the driver to take measures to avoid the occurrence of dangerous situations. It can not only guarantee driving safety but also is the basic technology to realize vehicle intelligence, and has gradually been recognized by automobile manufacturers, automobile R&D personnel and consumers.
Anti-collision system is an important function in ADAS. Its signal acquisition system can automatically detect the speed of the vehicle itself, the speed of the vehicle in front of the vehicle, and the relative distance between the vehicle and the vehicle in front through radar, laser, sonar and other technologies. Then, the control system judges whether the own vehicle is in a safe driving environment through the information collected by the signal system. When the control system judges that a collision may occur, it can remind the driver to perform operations such as safe driving through sound and light alarms. And when the system detects that the driver has not made any action, the control system will take automatic emergency braking and other operations to avoid collisions between the vehicle in front and the vehicle in front, effectively ensuring personal safety.
1. 4 Introduction to simulation tool MATLAB/Simulink
Nowadays, MATLAB simulation tools are widely used in all walks of life. The software was originally launched by MathWorks in the United States. It is used for commercial mathematical tools. It has very powerful functions, including control algorithm design, data visualization analysis and mathematical formula calculation. . There are mainly two kinds of simulation software used, namely MATLAB and Simulink modules.
.....
Need to complete this paper and simulation, private message