Research On Object Recognition And Localization Of Mobile Robot Based On Information Fusion

The research field of mobile robot has a hot topic currently,which is the research of target recognition and localization of mobile robot.The recognition and localization of target depends on the data information gathered by the vision sensor,the accuracy of the recognition and localization and the complexity of the algorithm is an important index to evaluate the quality of the machine vision system.In this thesis,we took a monocular robot system into consideration,combined with information of ultrasonic sensor to design a recognition and localization system based on the fusion of visual and ultrasonic sensor information.The system included image segmentation and recognition ? off-line calibration based on monocular camera ? monocular positioning of an image?ultrasonic distance measurement and so on.In summary,the work carried out and the results in this thesis are as follows:(1)Based on threshold segmentation of gray level images,a threshold segmentation method for color images based on HSV color space was proposed.Compared with the commonly used RGB color space,HSV color model for color expression was more in line with the definition of color,and color space was also uniform and less affected by light.The method is more suitable for the object of simple color,which is to be identified,and the situation of the whole environment which exists not much similar color disturbance.(2)For target matching recognition,a matching algorithm based on contour moment invariant template was proposed,according to the shape feature.Contour extraction of two valued image after segmentation,then matching.Compared with the traditional shape matching algorithm,this algorithm reduced matching pixels and saved computing time.(3)For the target localization,combined the ultrasonic sensor with the vision sensor,toestablish a localization model of multi-sensor fusion.And through the target localization experiment based on the NAO robot,to compare with the ideal of traditional plane constraint ranging model and a single dual image positioning algorithm.Experiments show that in the laboratory environment,the ultrasonic device of NAO robot can quickly detect the target and return distance value for single monocular localization model,and get a more rapid and accurate effect to compute the coordinate of target object.Compared with binocular vision system,monocular vision system the advantages of flexible action,simple structure and convenient calibration.Whether it is in the recognition test or localization experiment,we take not just a single information.In recognition,we combine two features of the color and contour.In localization,we take advantage of the information's fusion of visual and ultrasonic.In order to get a more precise result through the information making up for each other.In the subsequent,we can choose various sensor information to integrate,to do a deeper research.