【作者】 张娜；

【导师】 党宏社；

【作者基本信息】 陕西科技大学 ， 电力电子与电力传动， 2015， 硕士

【摘要】 在某些工业或者特殊领域中会用体积参数来对物体进行分类、识别等,这就涉及到计算不规则物体的体积,如煤矿传输带上的煤块、大型物料堆、挖掘的文物、人体器官等。本文通过非接触测量方式,利用计算机视觉技术,来实现不规则物体的体积测量,主要研究以下四个方面:(1)图像信息获取方法研究。图像获取装置包括两个环形光源,三个摄像头,其中摄像头与图像采集卡相连接,通过图像采集装置获得三幅图像,分别是左视图、右视图和俯视图,三幅图像为后续的体积测量做好前期的采集工作,本采集系统适合测量形状不是特别复杂,底部内陷较少且颜色均匀的不规则物体。(2)贝叶斯压缩感知与形态学成分分析的图像修复方法研究。图像的预处理包括灰度化、图像增强、图像修复、图像分割。采用平均法对图像进行灰度化,用同态滤波法对图像进行增强,用C-V模型法对不规则物体的图像进行分割。对于一幅图像,若其部分受到破坏,则对后期的体积测量造成了较大的误差,因此在测量前需要对破损的图像进行修复,采用贝叶斯压缩感知与形态学成分分析的图像修复方法,首先稀疏图像的结构和纹理部分,然后用贝叶斯压缩感知得到稀疏系数的分布函数,最后利用分布函数获得修复图像,通过比较SSIM值和PSNR值来验证其有效性,得出该方法修复效果较好。(3)明暗度恢复物体高度方法研究。利用明暗恢复形状的原理,通过分析物体高度值和图像灰度值之间的关系,建立二者的关系表达式。首先对灰度值进行归一化,剔除奇点对体积测量造成的影响,然后利用图像的灰度值来得到物体的高度值,利用该方法得到了石头和鸡蛋的高度,最后通过仿真恢复得到的石头的三维空间结构与石头实际的结构基本相似,其中冷色部分的高度值低,暖色部分的高度值高。(4)体积计算方法研究。利用明暗度恢复得到物体的高度值之后,主要是通过统计像素的高度值得出石头和鸡蛋的像素体积,然后利用像素当量值来获得不规则物体的实际体积。为了验证该方法的有效性,与采用排水法得到的鸡蛋和石头的实际体积进行比较,利用绝对误差和相对误差来分析测量结果的可行性。利用图像的明暗度对物体的高度进行恢复,通过与实际体积进行比较,最后计算出的体积相对误差基本在5%左右,在一定程度下能满足不规则物体的测量要求。更多还原

【Abstract】 Volume is used to classify, recognize in some industries or special fields, which relates to calculate the volume of irregular object, such as coals on transmission belt, large material stacks, cultural relics, human organs, etc. In this paper the volume of irregular object is obtained by computer vision technology through non-contact measurement. This paper mainly researches the following four aspects:(1)Acquiring method for image is researched. The image acquisition device comprises two annular lights, three cameras(left, right and top). Three cameras are connected with image acquisition cards in the computer.Three images are obtained by image acquisition device respectively, the left view image, right view image and top view image.This image acquisition is prepared for volume measurement. The acquisition system is suitable for the irregular shape is not particularly complex, and bottom invagination is less with uniform color.(2)An image inpainting method based on Bayesian compressive sensing and morphological component analysis is researched. Image preprocessing includes graying, enhancement, inpainting, segmentation. The image is grayed by average method, enhanced by homomorphic filtering method.Image segmentation use CV model. For a damaged image, it would cause a greater error, so before measuring volume the first important thing is need to repair the damaged image. To the question that the image of irregular object, an image inpainting method based on Bayesian compressive sensing and morphological component analysis is introduced. This method transforms the sparsity for structure and texture, then gets the posterior distribution functions of the two sparse coefficients through Bayesian compressive sensing. The inpainting image is obtained through distribution functions. The proposed method has a good performance compared with the traditional methods through SSIM values and PSNR values.(3)The recovery of the heights for irregular objects is researched. Analysis the relationship between height and gray through the principle of shape from shading, then the relationship is established. Gray values are normalized firstly to avoid the impact of singular points. Then heights are obtained through gray values, the heights of stones and eggs are obtained through this method. The recovery structure of stone is similar to the actual structure through simulation. Cool part means the height value is low. Warm part means the height value is high.(4)The method of calculating volume is researched. First the heights of stones are obtained. Pixel volumes of stones and eggs are calculated through the total height. The actual volumes are obtained through pixel equivalent values. The absolute errors and relative errors are used to analysis the measurement results. Use the actual volume of stones and eggs to verify the effectiveness of the method.The heights of irregular objects are recovered through shadings of images. Tested the results with the actual volume by drainage method, the relative errors are about 5% mostly, and it can meet the measurement requirements of irregular objects in a certain extent.更多还原