Research On Measurement Algorithm Of Processing Targets For Bamboo Splitting Machine Based On Machine Vision

With the increase of demand for bamboo, bamboo machinery automation has attracted people’s high attention, but the traditional bamboo industry machinery seriously restricts the development of bamboo industry. Thus bamboo industry machinery is imperative in the development of the high automation degree. With the development of computer and photoelectric technology, Vision measurement technology has wider application field and people have paid much attention to it. Vision detection system for bamboo splitting machine based on machine vision overcomes the deficiency of the traditional bamboo splitting machine, which has important theoretical significance and application value for the development of bamboo industry machinery.Regarding digital image processing and machine vision technology as the background and basing on the bamboo mathematical model theory, we respectively studied bamboo CNC splitting machine target segmentation, detection and tracking, and monocular vision measurement. We are based on machine vision to design visual inspection software system of Bamboo CNC splitting machine automation detection. By using digital image processing technology, we research bamboo face image preprocessing and segmentation algorithm, and face image of bamboo is extracted. We use machine vision technology to research moving object detection and tracking algorithm, so target detection and tracking of bamboo timber processing is realized. And we use monocular vision technology to research the camera model, calibration technology, which is based on image ranging algorithm and bamboo geometric parameter measurement of imaging theory.Finally we realize the application of the automatic detection and tracking and machine vision measurement technology in the automatic processing system about the numerical control bamboo splitting machine. In this paper, we obtain the following several important conclusions:(1) Based on machine vision technology, we have established cutting bamboo machine vision inspection system and have completed hardware and software design of Visual detection system. For software system design and development,we use Visual C++. The structure optimization, the function is all ready and man-machine interfaces are friendly. These can realize the automation measurement of split bamboo and improve the processing efficiency.(2) We systematically studied the bamboo mathematical model, the characteristics of three kinds of bamboo timber processing ways and the calculation method for bamboo cane volume and the utilization. Three kinds of bamboo timber processing ways are Four sides flat milling, Two sides flat and two sides arc milling and Four sides milling and three splitting processing. From the bamboo tube subdivision number, length and the small end diameter and so on several aspects, bamboo timber volume and the rate of material has carried on the analysis and comparison.(3) Image processing algorithm for visual inspection system is studied. Three kinds of image segmentation algorithms based on wavelet transform and color space face have put forward. Image segmentation algorithm based on wavelet transform respectively uses the non-sampling Contourlet transform and Gabor wavelet transform to extract image texture feature and realize the image segmentation experiments. Image segmentation algorithm based on color space makes use of Lab color space and overlap of bamboo color to design characteristics of the bamboo image segmentation, and a circle fitting method is used to the end of the bamboo calibration. The algorithm has achieved the very good segmentation effect.(4)Moving target detection and tracking algorithm of visual inspection system is studied. Gaussian mixture model moving target detection algorithm is proposed, which is based on improving frame difference. Research the insufficients of frame differential method and Gaussian mixture model running target detection algorithm to fusion the both, using the frame difference to distinguish different sports section of the video. And giving take different update strategy for different regions, Gaussian mixture model are used to get the foreground image, so as to realize the detection of moving targets. The moving target tracking algorithm based on improved Camshift is proposed. Research the algorithm, using more features be composed color, texture, gray gradient to build a joint probability distribution and realize the accurate tracking of moving object. In the bamboo motion target detection, we use background subtraction and frame difference to realize bamboo moving target detection. Camshift has on tracking the moving targets with bamboo. Kalman filter realizes the prediction of moving target and the target location. Algorithm improves the accuracy of the bamboo target tracking, and processes the shadow, target and the background of similar problems. The algorithm improves the stability of tracking and has good adaptability.(5) Measurement technology of visual inspection system is studied. We put forward object size information measurement algorithm based on monocular vision. Researching the camera model and calibration technology, and using the geometric relationships between a camera model, according to the imaging theory of measuring two feature points in the image and its distance to the camera at the midpoint, and then the size of the object was derived according to the geometric information. The algorithm has simple design ideas, fast calculation speed and easy to implement, which is the innovation for measure theory.