Research On Coordinate Control Of Biomimetic Dual PTZ Cameras System And Mobile Platform

Vision is an important method of mobile robot to detect external environment. How to quickly search and track a focus of attention in a wide range of surroundings has become a new research direction. This subject is supported by Project of Specialized Research Fund for the Doctoral Program named "Biomimetic ground perception and motion control of autonomous robot". This purpose of this paper is to build a platform of visual system with a set of independent pan-tilt cameras and present a framework of environment visual scaning,target searching and tracking which is inspired by our understanding of chameleon visual system.Firstly, according to the research status and the tendency of mobile robot visual system,the platform of Biomimetic visual system is proposed. In this system, PTZ camera has the character of pan, tilt, and zoom. For this consideration, the subject chose the IPC as the main control unit. Steering servo control card is chosen for the steering servo controlling.Lens controller communicated by 485 bus is chosen for lens controlling. The control system has to be able to control the platform, beside, the attitude, the position and zoom ratio of lens.Secondly, according to the existing vision platform, this paper applies coordinate transformation method,establishes kinematics model, and analyzes the relevant characteristics. The relationship between world coordinate and camera coordinate is derived by kinematics model. In addition, an approach of bottom-up visual attention is framed based on the theory of human visual attention. In this model, intensity, color and orientation which used as features attracted attention are resolved into multi-scale through Gaussian pyraminds and conspicuity maps of each channel are calculated. Focus of attention is detected by integration of each channel.Thirdly, an objective searching model is derived based on the chameleon visual system.The variable resolution alignment model is derived to fulfill the task of precisely alignment value point target. The turning optimization model which based on genetic algorithm is derived to improve the efficiency the mobile robot. In addition, an image tracking algorithm based on particle filter is used to trach the target which selected by visual attention algorithm.Finally, bottom-up visual attention model is testified by detecting the focus of attention on static picture. Besides, carry out experiments on hard ground and uneven sandlot, to testify the variable resolution alignment model, and the turning optimization model, at the same time, angle of each axis, image of each camera are fed back. Via comparing the target position and the center point of image, the image tracking algorithm is testified. In addition, the experiment also tests the stability and excitability of the control system.