本文選題：三維重建 + IMU和視覺(jué)融合��； 參考：《浙江大學(xué)》2017年碩士論文

【摘要】：在電商化的推動(dòng)下,腳型三維數(shù)據(jù)的獲取工作在鞋類(lèi)定做領(lǐng)域占據(jù)重要地位。傳統(tǒng)的計(jì)算機(jī)多視覺(jué)幾何算法對(duì)三維腳型進(jìn)行重建需要滿足各種各樣的條件,如要求腳上紋理明顯(要求數(shù)據(jù)獲取時(shí)穿帶特殊標(biāo)記的襪子等),極為不便,阻礙了腳型重建技術(shù)的普遍適用性發(fā)展。本文基于移動(dòng)終端對(duì)腳型數(shù)據(jù)的獲取問(wèn)題進(jìn)行研究,實(shí)現(xiàn)了一個(gè)自動(dòng)腳型模型生成系統(tǒng)。本文的主要工作有以下幾方面:1.由于純視覺(jué)的三維腳型重建算法對(duì)輸入圖像的紋理要求較高,魯棒性差。本系統(tǒng)在以純視覺(jué)重建算法為主導(dǎo)的同時(shí),融入了IMU數(shù)據(jù)估計(jì)圖像幀對(duì)應(yīng)的攝像機(jī)姿態(tài),當(dāng)視覺(jué)重建算法因?yàn)樘卣鼽c(diǎn)過(guò)少或其它原因失敗時(shí),用IMU預(yù)測(cè)的攝像機(jī)姿態(tài)作為當(dāng)前圖像幀的初始值,結(jié)合特征點(diǎn)、輪廓等已有數(shù)據(jù)的相互約束關(guān)系進(jìn)一步優(yōu)化攝像機(jī)姿態(tài)。2.由于噪聲等原因帶來(lái)的誤差,得到的腳型三維點(diǎn)云與真實(shí)腳型之間還存在一定的差異,特別是當(dāng)輸入圖像幀并沒(méi)有包含整個(gè)腳型時(shí),重建出的三維點(diǎn)云存在部分點(diǎn)缺失的情況。為了解決上述問(wèn)題,引入?yún)⒖寄_型模型作為參考,當(dāng)重建出的三維點(diǎn)云有缺失部分時(shí),先用參考腳型模型的數(shù)據(jù)進(jìn)行補(bǔ)充。然后對(duì)兩模型之間的對(duì)應(yīng)點(diǎn)距離、面片法向量、網(wǎng)格曲率、輪廓、骨架等一系列數(shù)據(jù)信息建立約束關(guān)系,調(diào)整重建點(diǎn)云中的三維點(diǎn)坐標(biāo)進(jìn)一步修復(fù)重建模型,最終得到完整的三維腳型網(wǎng)格模型。3.基于本文的研究成果,實(shí)現(xiàn)了一個(gè)基于移動(dòng)終端的三維腳型數(shù)據(jù)獲取系統(tǒng),該系統(tǒng)實(shí)現(xiàn)了用帶IMU傳感器的移動(dòng)終端獲取靜止?fàn)顟B(tài)下的腳型三維數(shù)據(jù)的功能,并通過(guò)測(cè)量腳型形狀的相關(guān)參數(shù)來(lái)衡量最終重建出的腳型的精確度。最后,文章對(duì)目前已有的工作進(jìn)行了總結(jié),并對(duì)后續(xù)工作做出了展望。
[Abstract]:With the promotion of E-quotient, the acquisition of three-dimensional data plays an important role in the field of shoe customization. Traditional computer multi-vision geometry algorithm for 3D foot reconstruction needs to meet a variety of conditions, such as requiring a clear texture on the foot (require data acquisition wearing specially marked socks, etc.), extremely inconvenient, It hinders the development of universal applicability of foot reconstruction technology. In this paper, an automatic foot model generation system based on mobile terminal is studied. The main work of this paper is as follows: 1. The 3D foot reconstruction algorithm based on pure vision requires high texture of input image, and its robustness is poor. The system takes pure visual reconstruction algorithm as the dominant method, and integrates IMU data to estimate the camera posture corresponding to the image frame. When the visual reconstruction algorithm fails because of too few feature points or other reasons, The camera attitude predicted by IMU is used as the initial value of the current image frame, and the camera attitude is further optimized by using the constraint relation of the existing data such as feature points and contours. Because of the error caused by noise, there is still a certain difference between the 3D point cloud and the real foot, especially when the input image frame does not contain the whole foot, the reconstructed 3D point cloud is partially missing. In order to solve the above problem, the reference foot model is introduced as the reference. When the reconstructed 3D point cloud has missing parts, the data of the reference foot model are used to supplement the model. Then the corresponding point distance between the two models, the normal vector of the surface, the curvature of the mesh, the contour, the skeleton and a series of data information are set up, and the 3D point coordinates in the reconstructed point cloud are adjusted to further repair the reconstructed model. Finally, a complete three-dimensional foot mesh model. 3. 3. Based on the research results of this paper, a 3D foot data acquisition system based on mobile terminal is implemented. The system realizes the function of using mobile terminal with IMU sensor to obtain three-dimensional data of feet in static state. The accuracy of the reconstructed foot shape is measured by measuring the relevant parameters of the shape of the foot shape. Finally, the paper summarizes the existing work and prospects the follow-up work.
【學(xué)位授予單位】：浙江大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP391.41

【參考文獻(xiàn)】

相關(guān)博士學(xué)位論文 前1條

1 高飛;基于多視幾何的三維腳型測(cè)量技術(shù)與系統(tǒng)[D];浙江大學(xué);2010年

，

本文編號(hào)：1859945