Study And Application Of Machine Vision In Non-contact Measurement

Machine vision measurement systems have the features of non-contact, real-time, on-line measurement and high accuracy; they have been developed fast and used widely. At first the definition, history and technical difficulty of machine vision are simply introduced in this thesis. Then the machine vision measurement technique and its status in china and abroad are also introduced. It is pointed out that speed and precision are the two key points of the application of machine vision measurement.For achieving the requirements of speed and precision, with application experiences, it is in detail discussed that how to choose suitable hardware and how to design the correct algorithms. The system hardware mainly includes illumination, camera, lens, frame grabber etc. And the image processing arithmetic mainly include filtering & image enhancement, image threshold segmentation, edge detection, sub-pixel edge detection, camera calibration, shape-based matching, etc.Two machine vision measurement systems are picked from all systems which are all designed and finished by myself. The two systems are in detail discussed in this thesis. It is introduced that how to choose hardware and design image processing algorithms based on real requirement.The first system is automobile saddle measurement system. It carries out inspection of circuit parts of saddle. The main functions are solder point inspection, position inspection, loop circle inspection, diode inspection etc. The median filter is improved in this system for better effect. And the difficult problem about segmentation of soldering tin is solved through combining geometry feature with histogram segmentation.The second system is copper cast measurement system. It is for getting rotated angle relative to correct orientation. Many models with different angle are built and the searching region is limited, the speed and precision of model matching are improved greatly.