| Currently,machine vision technology is widely used in intelligent manufacturing,reverse engineering,industrial automation and other fields,and the measurement technology based on 3D reconstruction has become an important research direction of machine vision.There are many technical means of 3D reconstruction.Among them,the 3D reconstruction system based on line laser scanning owns significant advantages such as high measurement efficiency,simple structure,easy maintenance and available to many scenarios.Since the 3D camera is the core component of the line laser 3D reconstruction system,this thesis makes a study on its key technology,which plays an important role of improving the 3D camera function as well as its performance.Based on the existing 3D camera prototype of the research team,the key technology of 3D camera software is studied,and the main contents are as follows:(1)The 3D camera system composition is introduced and its software key technologies are analyzed.First,the laser triangulation principle,system composition and 3D imaging process are introduced;then the hardware structure is analyzed,and the structure,performance and characteristics of the main processor Zynq-7000 and CMOS sensor PYTHON 2000 are briefly described,and the underlying hardware architecture is analyzed;then the software architecture is analyzed,and the driver framework and Gig E Vision protocol are described in detail;finally,the In-depth analysis of software key technologies which include laser stripe centerline extraction algorithm and ARM-based software key technologies.(2)The laser stripe centerline extraction algorithm is studied.Firstly,the characteristics of laser stripes and the influencing factors of centerline extraction are analyzed,and pre-processing methods such as image filtering and image segmentation are used;then the common centerline extraction algorithms are analyzed,including the polar method,grayscale center of gravity method,skeleton refinement method and Steger’s algorithm.Finally,based on the improved grayscale center of gravity method proposed by related scholars,the algorithm is further optimized by combining the characteristics that image noise usually conforms to Gaussian distribution,and the classical grayscale center of gravity method is used to obtain the initial estimate of the center point as the center of the circle to select a circular region,while the radius of the circle is related to the radius of curvature of the center line.The weighted distance in the normal direction is calculated based on the gray weighting,and the center point is corrected using this weighted distance so that the horizontal coordinate of the new center point is no longer an integer(column number),but a sub-pixel precision.This improved algorithm is targeted for application scenarios where the normal direction deviates significantly from the column direction of the stripe image,such as surface contours.(3)Implemented key ARM-related software technologies.The device driver of PYTHON 2000 CMOS image sensor was ported to enable the 3D camera to obtain laser stripe images;the camera embedded application software was optimized by adding a centerline extraction module and reserving an interface for the embedding of various algorithms,and the communication module was optimized;the remote firmware upgrade of the camera was designed and implemented by combining Hoffman coding and Gig E Vision protocol The QSPI FLASH + e MMC camera boot mode is implemented to solve the problem that the TF card boot mode is vulnerable to vibration.Finally,a 3D camera test system is built to systematically test the functions of the camera in terms of startup mode,firmware upgrade,driver porting,and centerline extraction in turn.Different algorithms are used to extract the center point coordinates,and the 3D contours are reconstructed and measured by the upper computer software.The results show that the improved grayscale center of gravity method outperforms the classical grayscale center of gravity method in surface contour reconstruction. |
PDF Full Text Request