Research On The Characteristics Of Robot Dual Sensor Information Extraction And Fusion

Alone with the development of the science and technology,robots have become more and more intelligent.The research on information fusion has made great contributions to robots.Different kinds of sensors can work in the way similar to human sensing organs.In the complex environment,the strong function of sensing environment will help robots to solve problems quickly and accurately.And information fusion technology is the foundation of robot's sensing environment ability.With the continuous improvement of intelligentization of robots,information from single sensor can't meet the demand that robots need to detect the environment.So it's necessary for us to study information extraction and fusion.Detailly stereovision can detect the shape,color and size of the target,but it can not get the information of the material,while the ultrasound sensors can detect the subtle differences of the target surface,therefore they can complement each other.In this paper,we have studied the characteristics of ultrasonic sensor and stereovision sensor.And we use some algorithms to fuse the information from them in order to identify the targets.After analysing the characteristics of ultrasonic sensor and stereovision sensor and many prevalent fusing algorithms,we determine to use DSP and ARM as core processors.The hardware consists of ultrasonic module based on DSP,stereovision module and information fusion module which are based on ARM.The software includes transmitting,receiving and processing ultrasonic wave,extracting the characteristic signals and fusing information.In order to test this system,we have designed experiments for each module.The experimental results show that the detection distance of the ultrasonic and the detection angle meet the demand.The stereovision module can extract target's distance,position,color and shape.The information fusion experiment is also carried out to test the result of target identification.Throughout the experiments,it is proved that the device can not only identify the targets,it can also update and expand the knowledge library.The system can judge whether the two sensors are collecting the same target information according to the position information,and we study the feature level fusion,which not only can improve the accuracy of target recognition,but also ensure the execution efficiency.So it can make contributions to blind guidance,detecting environment,path planning and auto-driving system.