【作者】 赵艳；

【导师】 刘教民；

【作者基本信息】 燕山大学 ， 计算机应用技术， 2013， 博士

【摘要】 三维重建是指对三维物体建立适于用计算机表示和处理的数学模型，是在计算机环境中对其进行存储、处理和分析其性质的基础，也是在计算机中表达客观世界的关键技术。在负载电路开关时，低压开关电器触头间会产生开关电弧，电弧是影响低压电器电寿命的主要因素。低压电器中电弧的生成形态、燃弧运动过程等对分析电器的开关能力、电器寿命以及电器是否会发生故障起着重要的作用。因此要改善低压电器产品的可靠性，就必须研究从燃弧开始到灭弧结束整个过程的动态变化，以及观察各种因素对电弧运动过程的影响。开关电弧燃弧过程非常短暂，不适合直接进行观察分析，因此对于开关电弧燃弧过程建立可视化研究模型，有助于对电弧燃弧机理进行深入研究、分析电弧特性，从而达到改善电器结构设计、改进电器材料、提高电器产品可靠性的目的。因此对本课题深入开展研究工作具有重要的理论意义和实用价值。首先，通过对现有技术的分析和比较，设计了基于高速CCD摄相机的低压交流接触器开关电弧图像采集系统，建立了系统的硬件电路和软件系统结构。对于采集到的电弧图像进行三维重建，针对三维重建过程中涉及到的边缘检测、特征点匹配、图像增强、摄像机标定、曲面拟合等问题开展了理论和实现技术的研究，并提出了相应的解决方案。其次，针对电弧图像的特点，利用形参均匀B样条公式，提出了一种可盈亏修正的边缘检测算法，利用形参均匀B样条公式对修正后的图像曲面进行拟合，通过计算获得图像特征点，该方法简洁，精度较高，便于实时处理。再次，提出了多特征融合的图像匹配算法，通过SIFT算法提出特征点，并对SIFT算法进行改进，通过在图像与高斯和卷积之前对图像放大一倍处理的方法提高了特征点的提取速度。其次将特征点的灰度、距离特征组合起来，共同描述一个特征点，根据这些特征计算特征点之间的相似度，获得更能描述图像的局部特征，利用该方法有效的减少伪特征点数，并精确地定位出特征点，有效地提高了识别效率。为保留电弧弧柱强特征，采用基于直方图的图像增强算法，取得了较好的图像增强效果。最后，本文提出了对于电弧图像的标定算法，建立了基于非均匀有理样条的曲面拟合算法，实现了低压电器开关电弧可视化仿真模型，对电弧开断过程实现了仿真模拟，通过计算机将灭弧室内电弧的运动过程显示出来，方便从不同角度进行观察。电弧可视化模型的应用对于电器设计水平的改善和提高大有益处。更多还原

【Abstract】 Three-dimensional reconstruction is to create mathematical model forthree-dimensional objects by computer representation and processing, this model is thebasis for storing, processing and analyzing the nature of object by computer, which is keytechnology to express objective world in computers.When loaded circuit switches, electric arc will happen between the contact of lowvoltage electrical apparatus, which is the main factor affecting the life of low-voltageapparatus. The generation and shape of low-voltage electrical arc and arcing motionprocess play importment roles in the analysis of switch capability, electrical life as well aselectrical failure. Therefore, to improve the reliability of low-voltage electrical products, itis necessary to research the dynamic change of the whole process from arcing start to thearcing end, and to research the impact of various factors on the process of arc movement.Because switch arc arcing process is very short, it is not suitable for direct observation andanalysis. To establish visualization model for arc arcing process helps to conduct in-depthresearch, analysis arc characteristics of the arc arcing mechanism, so as to achieveimproved electrical structural design, improved electrical materials and improved thereliability of electrical products. Therefore, to carry out research work on this subject hasimportant theoretical significance and practical value.Based on a comprehensive analysis of existing visualization techniques, thelow-voltage contactor switch arc image acquisition system was proposed, which wasbased on high-speed CCD camera. A system of hardware and software system structurewas designed to acquire arc image and then do three-dimensional reconstruction. Theoryand technology researches on three-dimensional reconstruction process such as edgedetection, feature point matching, image enhancement, camera calibration, surface fittingwas carried out and related solutions was proposedEdge detection is the basis for a variety of three-dimensional computer reconstructionalgorithm, impacting directly on the effect of the three-dimensional reconstruction. In thispaper, concerning the characteristics of the arc image, a profit and loss modified edgedetection algorithm was established based on uniform B-spline with shape parameter, which use uniform B-spline with shape parameter formula to fit the modified imagesurfaces, the image was obtained by calculating the feature points. This method wassimple, high precision, and easy for real-time processing.Feature point is an important feature of images, and plays an important role inthe field of three-dimensional reconstruction. This paper proposed a multi-featurefusion image matching algorithm. Feature points were obtained through SIFTalgorithm, the SIFT algorithm was improved as well. By a zoom doubled processingon the image before the Gaussian and the convolution to improve the speed of thefeature point extraction. Gradation and distance of the feature point were combinedtogether to describe a feature point, and to calculate the degree of similarity betweenthe feature points, these characteristics can describe localized feature better. Usingthis method, we can effectively reduce the number of false feature points, preciselylocate the feature point, and effectively improve the recognition efficiency. To retainthe characteristics of the arc column, the histogram-based image enhancementalgorithm was used; better effect on image enhancement was achieved.Finally, the arc image calibration algorithm was proposed, and based on non-uniformrational spline surface fitting algorithm, visual simulation model of the low-voltageelectrical switch arc was achieved, arc breaking simulation model was achieved. Theelectric arc interrupter movement process was displayed through computer simulation; itwas convenient to observe and analyze the arc from different angles. The application ofarc visual model can improve the electrical design and benefits the electrical apparatusindustries a lot.更多还原