The Design Of Anti-collision Early Warning System For Front Vehicle Based On Machine Vision And Lidar

As we all know,car traffic safety is a problem that has always existed and become more serious.In recent years,a large number of researchers and companies have invested a lot of energy in automotive active safety technology and smart car technology as well as unmanned technology.The purpose is to minimize the occurrence of unnecessary traffic accidents and to protect the personal and property safety of the public.The front vehicle collision avoidance warning technology researched in this paper is the cornerstone and core point of automotive active safety technology,so the research content of this paper has certain reference value and practical significance.Based on the principle of small hole imaging to establish the coordinate system,a new vehicle image recognition algorithm is developed by using Visual Studio plus OpenCV.Firstly,using the morphological knowledge of image processing and the shape characteristics of the vehicle,the threshold conditions are selected.And the target hypothesis area that satisfies the rectangular size feature of the tail of the vehicle;Secondly,the HAAR feature-based cascade classifier is used to re-identify the target hypothesis area,and the final target hypothesis area of the vehicle image recognition is obtained.A vehicle identification experiment was carried out by randomly capturing 200 images containing vehicle information.Experiments show that the image recognition algorithm has higher recognition rate and better real-time performance.Lidar is selected as another sensor for acquiring vehicle information,and sensor information is integrated with the camera sensor.The laser radar sensor ranging and scanning principle are analyzed,and the laser radar coordinate system is established.The preliminary screening processing of the lidar data is carried out,and the radar data is processed twice by the DBSCAN algorithm to obtain the core point.Then a new sensor information fusion method was proposed.Then the sensor fusion recognition experiment was carried out in the host computer.The experiment shows that the fusion recognition method is feasible,and the recognition accuracy of the target is higher than that of the single sensor.Considering the influence of environment and driver on the braking,an improved minimum safety distance model is proposed.Afterwards,the definition of early warning parameters is proposed,and different early warning mechanisms under different early warning parameter levels are designed.Then using STM32F103C8T6 as the core processor,the hardware part and software part of the warning module are designed.The hardware part mainly has an indicator warning module,a buzzer warning module and an LCD liquid crystal display module.The software part mainly includes the ranging speed subroutine,the calculation of the minimum safety distance subroutine and the execution of the alarm subroutine.Finally,indoor early warning experiments and early warning experiments on outdoor roads were carried out.The experiment shows that the early warning module of the front vehicle anti-collision warning system designed in this paper can work normally,timely and accurate early warning,and basically achieves the expected goal.It lays a solid foundation for the next step of active anti-collision brake control research,and provides relevant theoretical knowledge and practical experience.