Study On The Vision Sensor Technology For Robotic Applications

Vision sensor is a function-specialized vision system with image acquisition,image processing and data communication capabilities.It has become an indispensable key sensing method for intelligent industrial robots.The dissertation focuses on the application of vision sensor in industrial robots,and the key technologies and implementation methods of software and hardware architecture,image registration,firmware,graphical programming environment,and communication framework of vision sensor are studied.The major contributions are as follows:Feature-based image registration technology is the key to the perception function of robot vision system,such as object recognition,pose estimation,and so on.The improved algorithms based on the purification of matching pair are specifically studied.In order to solve the problem of low precision of ORB algorithm,an improved ORB algorithm is proposed to eliminate mismatches by neighboring ratio,bidirectional matching and triple angle cosine.In order to solve the problem of slow speed of D-Nets algorithm,an improved D-Nets algorithm based on matching quality purification is proposed.A general-purpose DNP algorithm combined with local features and global information is proposed and applied to a variety of registration algorithms including FREAK,SIFT,SURF,and ASIFT,and the experimental verification is carried out with Oxford Dataset.The process of camera calibration and hand-eye calibration is discussed.The target positioning is realized based on the contour profile feature.One-dimension and two-dimension barcode identification algorithm and its implementation are studied.The similarity edge measurement method is used to realize the identification of EAN-13 code.Based on the position detection pattern and the trapezoidal correction method,the position and correction of QR code is accomplished.The firmware architecture based on uClinux embedded operating system and graphical programming environment is designed and developed.The software implementation technology of vision application tools integration framework,job management,off-line simulation,and message transfer mechanism is studied.The basic framework of data interaction and the implementation of communication protocols are also discussed.The functional testing of off-line simulation and vision application tools of the graphical programming environment is performed.Based on the vision sensor prototype,vision aided alignment,two-dimensional code identification,and positioning and guidance application tests are performed on three different types of industrial robot platforms: Cartesian coordinate robot,mobile robot and articulated robot,and thereby the feasibility and practicality of the system are verified.