Color-shape Based Target Recognition And Localization With Omnidirectional Camera In Mobile Robots

The target recognition and localization of mobile robot is very important for the environment cognition,target tracking and path planning of mobile robot.Compared with the traditional vision system,the panoramic vision system has large imaging range and wide vision field,and the obtained image contains both location information and angle information,which is more conducive to the robot positioning of the target.The object of the study is mobile robot consisted of VS-C450U-200-TK omnidirectional cameras of VSTONE company.The article mainly discusses about the topic that how robots locate target in pictures by dealing with different colors and shapes.Firstly,the Ostu based image segmentation method is used to obtain the connected region of the image,and then filtering and labeling.The panoramic image collected by the mobile robot is pretreated.Then,the color histogram of the labeled region is extracted based on the HSV color space,and the color distance between the histogram is used as the measurement standard.The color similarity between the object and the target area is obtained.The aHash algorithm is used to calculate the Hash values of objects and objects in the labeled region.The Hamming distance between Hash values is used as a metric to obtain the shape similarity of objects and objects in the labeled region.Then,the similarity weighted model is established,and the weighted similarity between the object and the target in the marked region is calculated by color similarity and shape similarity,and the target is recognized from the panoramic image.After established the panoramic camera model and calibration,according to the corresponding image data obtained by the image distance–actual distance calibration,using three Hermite interpolation method of difference function between image distance and the actual distance.Finally,the distance formula between pixels based on the image from the target object and the robot was calculated,and the image distance calculated two azimuth angle and the actual distance,to achieve the use of color and shape information of the panoramic camera target recognition and location.In the first part of the thesis,several common panoramic imaging techniques are introduced,and the structure of panoramic vision system of mobile robot is analyzed,and the principle of single view catadioptric omnidirectional imaging is explained in detail.The second part describes the method of target recognition based on color and shape features,and gives a detailed description of color similarity,shape similarity,weighted similarity calculation methods and processes.The third part describes the method of target localization in detail,and explains the algorithm principle and process of panoramic camera calibration.In order to verify the performance of the target location method,two scenes are designed and the target location experiment is carried out.By comparing and analyzing the result of this method and other target location methods,the adaptability of this method is proved.The accuracy of the proposed method is verified by analyzing the target localization results of the mobile robot in stationary state.The stability of the proposed method is verified by analyzing the results of continuous target localization in mobile robot target tracking.