Research On Lane Recognization Technology Of Car Visual Navigation Based On ARM

Aiming at the safety problem of vehicle driving, this thesis researches a visionsystem of intelligent vehicle based on ARM9and Embedded Linux operation system.The overall framework of system is designed in this paper, the hardware has adoptedS3C2440microprocessor as the hardware which has the USB port connected with frontend camera, so that it can achieve monitor image acquisition. This thesis adopts Linuxoperation system for the embedded processor architecture as software platform which isopen source and scalable.Software design is completed which includes Linux kernel andfile system transplant, camera driver development, as well as the realization ofmonitoring application software.In this thesis，the road positioning algorithms based on image are researched toprocess the images. By combining many areas lane recognition system reference, firstanalyzes the current image-based preprocessing algorithm for lane recognition process,and then use the simulation experiment illustrates the complexity of the roadenvironment, the limitations of the traditional image preprocessing algorithm, Finallygenetic algorithm based on the complexity of the road environment to accuratelyidentify lane line algorithm. First, this article uses parabolic model to fit lane boundaries,and then uses the lane boundary and the road surface characteristics of gray-scaledifference and gradient difference to deal directly with grayscale images, increasing therecognition system in a complex road environment adaptability, but also improveefficiency of the system identification.Using genetic algorithm optimization searchcapabilities, to determine the lane line parabola of a,b and c parameters.Which giveseach parameter set the scope of the area of interest, to narrow search genetic algorithm,improve the identification accuracy and real-time performance.Finally, the simulation for lane recognition algorithms in a variety of complex roadconditions are verified, the results show the accuracy and timeliness to meet the systemrequirements.In order to optimize system performance, increase the real-time nature ofthe recognition system, from a shell script optimization, dynamic library optimization,code optimization layer optimization algorithm program analysis.At last, through theMATLAB simulation analysis, the establishment of auto visual navigation controlsystem has good control performance, can better realize the lane line identificationalgorithm is studied in the text, the control system shows good robustness.