工业物体表面三维视觉量测的关键技术研究

Evolutions and Key Techniques for 3D Object Surface Vision-Measurement in Industry

在总结国内外有关研究的基础上 ,提出了自己的研究思路 ,并对缺乏纹理目标的人工纹理产生和影像获取方式进行了研究。对圆目标影像的分割、探测、定位等算法以及利用自检校光束法平差方法解求物方空间坐标、带权观测值的权值给定等作了试验研究。给出了一个解算孔群孔心坐标的应用实例 ,物方控制点的检查点坐标中误差小于± 0 .2mm。

Digital close range photogrammetry system and machine vision are widely used in production control,quality inspection,and new production development in industry.The main aim is to provide precision 3D coordinates or reconstruction of an object surface and give an expression to an object shape.Current research findings indicate that calibration of digital camera has some problems in high precision measurement.The factors of causing image distortion have not been fully clear.Calibration model needs to be optimized for some application.Validity and adaptability of calibration need boosting.Detection,recognition and auto coding of artificial marks and coding marks and epipolar techniques are effective methods for coding of homonymous point in different images.Sub_pixel measurement precision is above 1/100 pixel under some condition.Our research focuse on automatic or semi_automatic measurement of object surface,and image match arithmetic.The arithmetic from single point matching to multi point matching and to multi photo matching is an important evolution. This paper reviews and analyzes systematically the key techniques for 3D object surface vision_measurement in industry,discusses and sums up new methods,new theory,new problem,and distortion parameter model of images,precision level in digital camera calibration.Several main artificial marks used in non_texture or poor_texture object surface measurement are introduced.The paper analyzes comparatively main methods for sub_pixel measurement and it can gain precision level in present,and indicates important evolutions in automatic and semi_automatic measurement,essential arithmetic or robust arithmetic for obtaining approximate coordinates of marks points in an object surface measurement.The paper also points out the importance of knowledge_dig and its application in mark recognition and detection,and emphasizes integrative techniques in vision_measurement system. In the end,a pilot study and a result of practice application are given.Research proved that the thought and methods about image segmentation,detection and location of circle marks were effective and valid. A propriety weight value for adding parameters,control points and oriented elements in self_calibration bundle adjustment are advantageous to gaining high precision of space coordinates.The R M S of check points of control point is ±0.2 mm.