基于灭点优化的单幅图像三维重建
|
摘要
单幅图像的三维重建避免了基于多幅图像重建的特征匹配的问题,是三维重建研究领域的一个热点。通过对单幅未标定图像进行三维重建,灭点的精度十分重要。提出一种利用直线的参数信息的灭点检测算法得到有效的直线,优化了灭点。首先对图像进行处理,提取出图像中的长直线,分析直线特征,对不同方向的直线进行分组,再根据各方向的直线满足线性分布关系,利用改进的回归算法获取直线参数的线性模型,剔除误差直线,再利用最小二乘法解算灭点。得到精确灭点后,根据灭点的性质获得摄像机的内外参数。交互获得最少的图像二维点,通过计算求得的摄像机内外参数和物体本身的几何特征计算相应的三维坐标,最后进行目标物体的三维重建。此方法有效剔除了无效直线的干扰,提高了灭点精度,以包装盒为例,重建了三维模型并且误差控制在2.5%以内,符合三维重建精度要求。
The 3 D reconstruction of single image avoids the problem of feature matching based on multi image reconstruction. It is a hot spot in the field of 3 D reconstruction. Based on the 3 D reconstruction from single uncalibrated image, vanishing point precision is very important. Using a linear parameter information of the vanishing point detection algorithm can effectively optimize the vanishing point line. Firstly, the images are processed to extract the length of the lineand linear features.The straight lines in different directions are grouped according to the direction of the line. Then according to the linear distribution relationship satisfied by the straight line in each direction, the improved regression algorithm is used to obtain the linear model of the linear parameter. The error of straight lines is eliminatedand the vanishing point is calculated through the least square method. To get the accurate vanishing point, the camera intrinsic parameters according to the nature of the vanishing point can be obtained. The minimum number of two-dimensional points of image is obtained by interaction. The corresponding 3 D coordinates are obtained by calculating the intrinsic and external parameters of camera and the geometric characteristics of objects themselves.Finally, the 3 D model of target object is reconstructed. This method effectively eliminates interference from invalid line and improve the accuracy of vanishing point. Taking the packing box as an example, the 3 D model is reconstructed and the error is controlled within 2.5%, which satisfies the requirement of 3 D reconstruction accuracy.
The 3 D reconstruction of single image avoids the problem of feature matching based on multi image reconstruction. It is a hot spot in the field of 3 D reconstruction. Based on the 3 D reconstruction from single uncalibrated image, vanishing point precision is very important. Using a linear parameter information of the vanishing point detection algorithm can effectively optimize the vanishing point line. Firstly, the images are processed to extract the length of the lineand linear features.The straight lines in different directions are grouped according to the direction of the line. Then according to the linear distribution relationship satisfied by the straight line in each direction, the improved regression algorithm is used to obtain the linear model of the linear parameter. The error of straight lines is eliminatedand the vanishing point is calculated through the least square method. To get the accurate vanishing point, the camera intrinsic parameters according to the nature of the vanishing point can be obtained. The minimum number of two-dimensional points of image is obtained by interaction. The corresponding 3 D coordinates are obtained by calculating the intrinsic and external parameters of camera and the geometric characteristics of objects themselves.Finally, the 3 D model of target object is reconstructed. This method effectively eliminates interference from invalid line and improve the accuracy of vanishing point. Taking the packing box as an example, the 3 D model is reconstructed and the error is controlled within 2.5%, which satisfies the requirement of 3 D reconstruction accuracy.
引文
[1] 陈志杨,罗飞.基于WebGL的Revit三维建筑模型重建[J].浙江工业大学学报,2016,44(6):608-613.
[2] 管秋,金俊杰,张剑华,等.基于最优RANSAC算法的非增加式多视图三维重建[J].浙江工业大学学报,2015,43(5):473-478.
[3] 潘翔,童伟淮,许洪涛,等.用户约束下的三维动画模型序列一致性分割[J].浙江工业大学学报,2015,43(4):420-424.
[4] 刘培珺,石教英,孙鑫.基于平面的建筑物表面模型重建算法的研究[J].计算机辅助设计与图形学报,2004,16(8):1045-1050.
[5] 余烨,刘晓平,,徐伟,等.面向建筑物重建的相机标定方法研究[J].图形学报,2012,33(4):76-82.
[6] CRIMINISI A, REID I, ZISSERMAN A. Single view metrology[J]. International journal of computer vision, 2000, 40(2): 123-48.
[7] 钱晶晶,余烨,刘晓平.基于几何约束的单幅图像三维重建方法[C]//计算机技术与应用进展学术会议论文集(下).合肥:中国科学技术大学出版社,2008:734-738.
[8] NIAN J, PING T, CHEONG L F. Symmetric architecture modeling with a single image[J].SIGGRAPH Asia, 2009, 59(3): 207-232.
[9] TARDIF J P. Non-iterative approach for fast and accurate vanishing point detection[J]. IEEE international conference on computer vision, 2009, 30(2): 1250-1257.
[10] XU Y L, OH S, HOOGS A. A minimum error vanishing point detection approach for uncalibrated monocular images of man-made environments[C]//Computer Vision and Pattern Recognition. Oregon:IEEE, 2013:1376-1383.
[11] 王思洁,方莉娜.基于结构化场景的单幅图像建筑物三维重建[J].地球信息科学,2016,18(8):1022-1029.
[12] 张桂梅,储珺,王丽.利用直线参数信息的建筑物灭点检测方法[J].计算机应用,2013,33(2):515-538.
[13] CIPOLLA R, ROBERTSON D. 3D models of architectural scenes from uncalibrated images and vanishing points[C]//Proceedings of the IEEE International Conference on Image Analysis and Processing. Chicago:IEEE, 1999:824-829.
[14] 李华.室内场景照片中的直线与灭点检测算法研究[D].西安:西安电子科技大学,2014.
[15] 王建,曾广宇.基于单幅图像的摄像机标定[J].微计算机信息,2010,26(7):214-216.
[2] 管秋,金俊杰,张剑华,等.基于最优RANSAC算法的非增加式多视图三维重建[J].浙江工业大学学报,2015,43(5):473-478.
[3] 潘翔,童伟淮,许洪涛,等.用户约束下的三维动画模型序列一致性分割[J].浙江工业大学学报,2015,43(4):420-424.
[4] 刘培珺,石教英,孙鑫.基于平面的建筑物表面模型重建算法的研究[J].计算机辅助设计与图形学报,2004,16(8):1045-1050.
[5] 余烨,刘晓平,,徐伟,等.面向建筑物重建的相机标定方法研究[J].图形学报,2012,33(4):76-82.
[6] CRIMINISI A, REID I, ZISSERMAN A. Single view metrology[J]. International journal of computer vision, 2000, 40(2): 123-48.
[7] 钱晶晶,余烨,刘晓平.基于几何约束的单幅图像三维重建方法[C]//计算机技术与应用进展学术会议论文集(下).合肥:中国科学技术大学出版社,2008:734-738.
[8] NIAN J, PING T, CHEONG L F. Symmetric architecture modeling with a single image[J].SIGGRAPH Asia, 2009, 59(3): 207-232.
[9] TARDIF J P. Non-iterative approach for fast and accurate vanishing point detection[J]. IEEE international conference on computer vision, 2009, 30(2): 1250-1257.
[10] XU Y L, OH S, HOOGS A. A minimum error vanishing point detection approach for uncalibrated monocular images of man-made environments[C]//Computer Vision and Pattern Recognition. Oregon:IEEE, 2013:1376-1383.
[11] 王思洁,方莉娜.基于结构化场景的单幅图像建筑物三维重建[J].地球信息科学,2016,18(8):1022-1029.
[12] 张桂梅,储珺,王丽.利用直线参数信息的建筑物灭点检测方法[J].计算机应用,2013,33(2):515-538.
[13] CIPOLLA R, ROBERTSON D. 3D models of architectural scenes from uncalibrated images and vanishing points[C]//Proceedings of the IEEE International Conference on Image Analysis and Processing. Chicago:IEEE, 1999:824-829.
[14] 李华.室内场景照片中的直线与灭点检测算法研究[D].西安:西安电子科技大学,2014.
[15] 王建,曾广宇.基于单幅图像的摄像机标定[J].微计算机信息,2010,26(7):214-216.