Algorithmic Research Of Automatic Cloth-cutting System

There is an urgent demand in cloth industry that is to perform cuttings on batches of cloth under certain rules. In the past, the job was performed by handcraft. However, due to different understanding of the same rule, manual work can hardly be satisfactory. Modern industrial manufacturing system requires fast and precise cutting of tons of cloth under strict rules, which brings birth to the Automatic Cloth-cutting System. The system acquires the visual information of cloth through CCD camera, and extracts the primary cutting lines using image processing algorithms, and then restores the geometrical information of cutting lines through camera calibration, and finally passes the information onto cutting sub-system of ACS to perform the assignment of cutting. There are two important algorithms in ACS: camera calibration and image processing algorithms that is dedicated to extract cutting lines of cloth. The following part of the thesis is organized as follows:Chapter 1: Introduces the background and kernel part of ACS, points out the main task of the thesis, and makes a brief introduction to the main two algorithms studied here: camera calibration and graph cut.Chapter 2: Looks into the problem of camera calibration, and introduce the status quo of the topic. According to the traits of the system, designs a new approach of camera calibration with advantages of avoiding the potential mismatch between analytical imaging model and camera, and of remarkable precision. Chapter 3: Designs the method to extract cutting lines on "contour-pattern" cloth automatically using maze-search algorithms and threshold method in graph cut. Chapter 4: Designs the method to extract cutting lines on "thin-line-pattern" cloth using maximal likelihood theory.Chapter 5: Summarizes all achievements of the dissertation, reviews points of innovation and defects, and gives the direction in future work.