Research On The Technologies Of Robust Guidance In Vision-guided AGV Under Complex Illumination Conditions

In order to solve the problem of Path Extraction for a vision-guided AGV under complex illumination conditions,this paper summarized the existing technology of vision-guided AGV,researched the extraction and the accurate estimate of path in visual guidance technology.Firstly,an adaptive image illumination partitioning and threshold segmentation approach based on a model of illumination and color was proposed for path extraction in the field of view for a visionguided AGV under complex illumination conditions.Firstly,the relation between light illumination and image brightness was analyzed,and the correlation model of illumination and color was built by measuring color distribution with respect to different illumination in images under complex illumination conditions.Secondly,the image of a guide path was partitioned into different illumination regions based on the support vector machine(SVM).Then the image of low-illumination region was enhanced in the space of RGB color in order to retrieve the color information of the guide path,and the image of highillumination region was processed by differentiating chrominance components of Cb and Cr in order to suppress the common-mode luminance interference and then an adaptive threshold segmentation method was performed for different illumination regions.Finnally a path extraction method based on particle swarm optimization(PSO).A large number of experimental results showed that this path extraction approach had the high adaptability to complex illumination when recognizing the guide path in the vision field with both high-reflective and dark-shadow regions caused by the environment illumination.Secondly,optimal estimation based on unscented kalman filter(UKF)was applied to robust guidance of vision-guided AGV.The nolinear AGV kinematics model which containted nosie was established as the system state model.The posture state was estimated using UKF taking the output value of visual recognition sensor as measurement.The robustness of vision-guided system was further improved by using the above method.Then,paper based with the technologies of path extraction and robust guidance under complex illumination conditions,developes the omnidirectional vision-guided AGV and does a large number of experiments with the AGV,experimental results show that the above methods is feasible and effective.Finnally,the main work done in this paper was summarized,and then the further work and the research directions in the future were discussed.