發(fā)布時(shí)間：2018-11-15 08:39

【摘要】：針對(duì)服裝動(dòng)畫中人體運(yùn)動(dòng)與服裝變形的關(guān)系,首先對(duì)服裝動(dòng)畫中人體運(yùn)動(dòng)特征和服裝變形特征進(jìn)行定義和提取;其次建立多套服裝在多種人體運(yùn)動(dòng)下的高精度服裝動(dòng)畫實(shí)例數(shù)據(jù),并基于4種機(jī)器學(xué)習(xí)模型從實(shí)例中學(xué)習(xí)二者之間的關(guān)系.對(duì)不同學(xué)習(xí)模型進(jìn)行實(shí)驗(yàn)的結(jié)果表明,由于服裝動(dòng)畫中人體運(yùn)動(dòng)與服裝變形之間存在較強(qiáng)的相關(guān)性,因此利用人體運(yùn)動(dòng)可以較為準(zhǔn)確地預(yù)測(cè)服裝的變形分布;相比BP神經(jīng)網(wǎng)絡(luò),廣義回歸神經(jīng)網(wǎng)絡(luò)和支持向量機(jī)等方法,隨機(jī)森林模型可以更有效地獲取人體運(yùn)動(dòng)與服裝變形的關(guān)系,最終誤差可控制在理想范圍內(nèi).
[Abstract]:In view of the relationship between human body movement and garment deformation in clothing animation, the characteristics of human body motion and clothing deformation in clothing animation are defined and extracted firstly. Secondly, the high precision clothing animation case data of multiple sets of clothing under various human motion are established, and the relationship between them is learned from the examples based on four machine learning models. The experimental results of different learning models show that because there is a strong correlation between human body motion and garment deformation in clothing animation, the deformation distribution of clothing can be predicted more accurately by using human body motion. Compared with BP neural network, generalized regression neural network and support vector machine, the stochastic forest model can obtain the relationship between human body motion and garment deformation more effectively, and the final error can be controlled within an ideal range.
【作者單位】： 華北電力大學(xué)控制與計(jì)算機(jī)工程學(xué)院;中國(guó)科學(xué)院計(jì)算技術(shù)研究所移動(dòng)計(jì)算與新型終端北京市重點(diǎn)實(shí)驗(yàn)室;
【基金】：國(guó)家自然科學(xué)基金項(xiàng)目(61300131,61532002,61272322) 裝備預(yù)研基金(9140A15030115DZ08042) 國(guó)家“八六三”高技術(shù)研究發(fā)展計(jì)劃(2015AA016405)
【分類號(hào)】：J218.7;TP391.41
，

本文編號(hào)：2332786