【作者】 楚万秀；

【导师】 戴惠良；

【作者基本信息】 东华大学 ， 机械制造及自动化， 2008， 硕士

【摘要】 数字近景摄影测量因其良好的可扩展性和强大的功能已逐渐成为目前逆向工程中最适用于工业现场测量的一种技术,近年来广泛应用于汽车车身及内外饰、飞机外身及桁架、大型天线等产品的设计制造过程中。随着大尺寸及结构复杂物体测量任务的增多,现有测量系统编码标志(也称为编码标志)总数的不足已经制约了其整体测量优势的发挥。扩展编码数目、提高数据处理能力,已经成为保证测量精度和工作效率的迫切需要。本文研究了编码标志的设计思想和解码策略,提出了一套新的编码设计改进方案和解码算法,研究了其信息自动检测过程中的图像处理、图像分析、信息提取加工等技术,并对实际处理中出现的问题给出了修正方案和校正模型。首先,对近景工业摄影测量的现状进行介绍,在分析现有的点分布式和同心环式两类编码标志点的特点的基础上,提出了一种新的编码标志设计方案。方案以一种非常实用的同心环式编码标志点为原型,添加了编码起始标志设计。这种方案既克服了旋转不变性造成的编码总数锐减的缺点,又发挥了其形式简单易于识别的优点,大大扩充了编码标志点的数目,满足了大尺寸和复杂物体的测量需求。然后,研究了国内外实用标志的解码方法,结合本课题设计编码标志点的特点,运用图像处理、图像分析、模式识别等领域的知识,设计了相应的解码算法,实现了解码的高效率和高准确性。在模拟编码图像的试验中取得了编码标志识别率、解码正确率100%的理想效果。最后,对实际场景拍摄实验图像进行研究,对实际拍摄图像解码过程中出现的错误进行分析。针对“起始线”与周围元素粘连的情况,提出了先腐蚀后膨胀的有效解决方法;针对拍摄角较大、拍摄距离较远时编码环段变形严重造成的错误识别情况,充分发掘相机成像模型所包含的数据关系,推导建立了整幅图像的校正模型。“起始线”粘连解决方法和校正模型在实际图像的处理中取得了非常良好的效果,证明了这种方法的有效性和校正模型的合理性。本课题提出的设计方案大大扩展了编码标志总数,拓展了编码标志的设计思路;设计的解码算法具有准确率高、效率高的优点;建立的全局校正模型解决了一类摄影成像的共性问题,在解码过程中取得了良好效果。各项实验数据结果证明课题的研究成果是完全可以满足工程中的实际需要的。此外,本课题建立的全局校正模型也适用于解决近景摄影测量领域中的同类问题,同时可作为图像处理领域相关问题研究的参考。更多还原

【Abstract】 In the field of reverse engineering, digital close-range photogrammetry is emerging to be the most extensible and powerful technique for industrial on-site measurement. It is being widely used to meet the requirement of design and manufacturing of products like automobile body & inner decoration, airplane body & skeleton, large antenna, etc. However, as tasks for measuring large and complex objects become much more common today, a shortage of the amount of coded target in existing systems has affected the advantage of this technique to implement overall measurements. It needs to find the solution which could be able to both expand the amount of coded target and enhance the capabilities of data processing to ensure the accuracy and efficiency of the system.In this thesis, the design concept of the coded target together with decoding strategies is represented. An improved design of coded target scheme and its corresponding decoding algorithm are put forward. For this purpose, techniques for auto information processing such as image processing, image analysis, and pattern recognition are researched. Furthermore, a rectification solution and a correction model are provided to solve problems related with real scene experiments.Firstly, the state of arts of industrial close-range photogrammetry is introduced. Based on the analysis of the two main types of coded target schemes, a new type of scheme is invented in order to fulfill the mission. The new scheme could be considered as upgrading the concentric ring type scheme by a start sign. Reserving all merits of the original scheme, this "starter" eliminates the effect of "rotation distinction" which is one main reason that limits coded target amount.Secondly, the decoding strategy and decoding algorithm are designed, and many tests of simulated ideal images are made. All related theories and techniques such as image processing & analyzing, pattern recognition are discussed. Decoding process produces perfect results in these tests.Finally, the decoding process in real scene image experiments is developed and tested. There are two problems during the test. One is that some starter sign may stick to surrounding region which makes decode fail. A 2-step rectification solution is designed to fix it. The other problem is that the targets of far and tilted photographing position are difficult to be decoded. For this, a correction model is established to process the image by utilizing projective transform model and its data relation. Both the rectification solution and the correction model are proved to be very effective and reasonable by tests.In conclusion, the entire solution proposed by the author is proved to be effective and practical. The amount of coded target is considerably enlarged; the decoding process is accurate and efficient; the rectification solution and the correction model are effective to handle the un-ideal data of real images. Moreover, now that the correction model is oriented to situations wherever projective transformation is involved, it could be used to solve similar problems in close-range photogrammetry and image processing.更多还原