基于三维标识物的医用增强现实跟踪注册方法研究

发布时间：2018-04-15 15:36

本文选题：增强现实 + 标识物设计　；参考：《吉林大学》2017年硕士论文

【摘要】：医疗领域因其高风险,不可重复的特点,迫切的需要更多的术前指导和术中辅助,因此本文针对增强现实在医疗辅助手术中的应用,引入了基于视觉的三维注册技术,将三重建后的虚拟模型与真实场景进行实时配准,从而更加直观的显示手术位置。为了提高虚实配准的效率和速度,本文设计了一个可在任何角度都能被检测到正六面体的三维标识物用来跟踪注册,重点研究了标识物的设计、识别、位姿估计算法,并将其应用在虚实配准实验中,实现了虚实模型的融合显示。实验证明了标识物的识别跟踪效果良好。本文的研究具体内容分为以下几点:1.在医疗辅助手术背景下,针对真实颅骨模型的三维特点和实验场景单一,而平面标识物具有在极限视角下突然失效的不足,设计了一种基于海明编码的正六面体三维标识物,不同面是编码不同的黑白色快。设计程序算法实现了标识物的检测和定位,识别得到了标志面的ID信息。2.针对三维标识物的不同识别情况,采用Pn P多点透视结合POSIT算法求得摄像机的位姿,为了尽可能多的识别三维标识物上的平面标志,在场景范围内对不同大小的标识物进行实验,获得了标识物的尺度优化函数,确定标识物边长为8cm。采用MC算法对颅骨模型CT数据三维重建获取了虚拟模型并渲染优化。3.对本文提出的跟踪注册方法进行实验验证,对摄像机进行标定获取内参数,采集的28幅图像标定的平均误差为0.38像素。将虚拟模型导入场景与真实颅骨模型配准显示,虚实模型辅助支架及上面的小球融合效果良好,保证了虚实配准的准确性。本文设计的三维标识物系统较为有效的实现了三维注册,识别速度在60ms,保证了识别的快速性。在场景范围长、宽、高为60cm×50cm×40cm,真实颅骨模型尺寸为20cm×14cm×14cm条件下,选取的标识物可使识别标志面占可视标志面个数百分比达到86.67%,保证了跟踪的有效。最后成功的将该标识物应用于虚实配准中。本文的研究成果为医用增强现实的研究提供了可行的思路。
[Abstract]:Because of its high risk and non-repeatable characteristics, the medical field urgently needs more preoperative guidance and intraoperative assistance. Therefore, aiming at the application of augmented reality in medical assistant surgery, a visual 3D registration technique is introduced in this paper.The virtual model is registered with the real scene in real time, so that the location of the operation can be displayed more intuitively.In order to improve the efficiency and speed of virtual and real registration, this paper designs a 3D marker which can detect hexahedron from any angle to track registration.It is applied to the virtual reality registration experiment to realize the fusion display of the virtual reality model.Experimental results show that the identification and tracking effect of the marker is good.The specific content of this paper is divided into the following points: 1.In the context of medically assisted surgery, a hexahedron 3D marker based on Hemin coding was designed to solve the problem that the three-dimensional characteristics of the real skull model and the experimental scene were single, while the plane marker had the deficiency of sudden failure in the limit angle of view.Different faces are coded with different black and white colors fast.The algorithm of design program realizes the detection and location of the mark, and obtains the ID information of the mark surface. 2.In view of the different recognition situation of 3D marker, using PN P multi-point perspective combined with POSIT algorithm to get the pose of camera, in order to recognize as many plane marks on 3D marker as possible,The scale optimization function of the marker was obtained and the side length of the marker was determined to be 8 cm by experiments on the different sizes of the marker in the scene range.The virtual model is obtained by using MC algorithm to reconstruct the CT data of skull model and the rendering optimization. 3. 3.The tracking registration method proposed in this paper is verified experimentally. The camera is calibrated to obtain internal parameters. The average error of the 28 collected images is 0.38 pixels.The virtual model was introduced into the scene and the real skull model was registered and displayed. The virtual solid model assisted the scaffold and the pellet above were fused well, which ensured the accuracy of the virtual real registration.The 3D marking system designed in this paper has realized the 3D registration effectively, and the recognition speed is 60 ms, which ensures the rapidity of recognition.Under the conditions of long, wide, high 60cm 脳 50cm 脳 40 cm high and 20cm 脳 14cm 脳 14cm of the real skull model, the selected marker can make the identification mark surface reach 86.67% of the visible mark surface, which ensures the effectiveness of tracking.Finally, the mark is successfully applied to virtual registration.The research results of this paper provide a feasible way for medical augmented reality research.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TP391.9

【相似文献】