Research And Application Of Visual Localization Technology Based On Machine Vision

Now, the main task of the industrial robot is running the predetermined operation instruction through assigning robot’s position in the working space, attitude and trajectory, and the operation sequence and time. When the robot’s working environment changes, its abilities about perception and adaptation to the environment are poor. Therefore it lagged the development of the field of industrial robots. Thus robots’ vision control technology has risen in recent years. There is no need to teach program offline for the robots’ trajectory and the actions anymore, so the introduction of machine vision technology can greatly improve industrial’s production efficiency and degree of production’s automation.This paper discusses plug-in robot’s visual localization algorithm and proposes a new visual localization algorithm which is based on Hough transform, aiming to obtain accurate position coordinate of the PCB board for the plug-in process of plug-machine. Preprocessing the image plays an important role in locating the work sites accurately. Using Gaussian filter to smooth the image can suppress the noise of the image; Adopting adaptive threshold method to binary transform the image can reduce the impact of environmental factors such as light intensity; The opening operation in morphological smoothes the objects’ contour, breaks image’s narrow connection and removes the small protrusions, making the details of the work sites in the image more obvious; Canny edge detection operator can position the edge better, maintain a weak edge extraction and reduce the pseudo-edge points; Propose a Hough transform algorithm based on image pyramid to detect circle. Compared with the traditional Hough transform algorithm, the improved algorithm reduces the non-zero points, replaces the complex three-dimensional accumulator with a two-dimensional accumulator, and decreases the time complexity and space complexity. Finally, calculating every board’s work sites’ coordinate through affine transformation can reduce the image distortion due to the position error, in order to control manipulator’s plug-in action.This paper presents a proposal to implement the software of the visual localization system. The proposal analyzes the main modules of visual Localization system and the relationships between them, develops plug-in robot’s visual localization system which is based on computer vision, and implements a Hough transform algorithm which is based on image pyramid to accurately identify the work sites and get the coordinate values. The experimental results from the industrial site show that this system can locate the PCB’s work sites accurately and can be used in the plug-in process of plug-machine perfectly with accuracy, stability, efficiency.