Research On Key Technology Of Vision Measuring System For Round Steel

The operation of steel rolling is the key procedure in the steel industry. Itprovided important basic materials for all sectors of the national economy, having aprofound effect on the manufacture level of large outfits in transportation, aviation,military projects, et al. The research of a practical online measuring device of largesteel rolling is important in a meaning that it could achieve the real time detection ofthe geometry, shape errors and surface defects of large steel rolling. The precisionmanufacturing capabilities of large steel rolling could be enhance further.Combined with the existing online measurement technologies of rolling, thispaper designed a multi-vision sensor measurement system which based online-structured light vision measurement methods. The system is designed formeasuring round steel which diameter is between200mm-900mm, and the requiredaccuracy is0.5mm. The key techniques of measurement system were studied torealize measurement of the geometry of large round steel. This paper mainly coversthe following aspects:1. This paper has introduced the principle of structured light measurement andproposed a plan to build a line-structured light vision measurement system.2. The paper has proposed a kind of multi-vision sensor calibration method basedon a three-dimensional shape of the target, which can solve the key issue formulti-vision sensor calibration. The method uses the principle of invariance of crossratio to get multiple calibration points on the light plane, realizing the local and globalcalibration of vision sensor synchronously. Just collected calibration points once thatcould complete the calibration of four sensors.The calibration method has higheraccuracy and is easier to operate, which means more convenient and more efficient.3. The main image feature extraction method involved in the measurementprocess was introduced in detail, including location of the target feature circle`s center,extraction of laser center and positioning algorithm of image sub-pixel. Throughextraction of characteristics of circular array, we identified the target under differentcamera views and located the sub-pixel in the center of oval. Based on adaptivewindow curve fitting algorithm, the center of laser was accurately located. It was proved by experiments that this method is robust and measurement requirements aremet.4. The design has completed the experimental verification relative to the systemmainly by choosing the large size rolling as the main measuring object. It also hasanalyzed the main error sources of the vision measurement system and has given outthe process on how the calibration error affected the3D reconstruction errors.