手绘服装草图的检索技术研究

A Retrieval Technology of Hand Drawn Sketch

以手绘形式在屏幕上勾勒出草图来获取信息、表达意图的方式因具有更加便捷与自然的特点,从而使以触控作为主要人机交互方式的移动设备逐渐成为主流.经试验对照组定量分析,改变传统SURF'算法,提高了特征匹配检索效率.利用空域平滑处理的均值算法使图像平滑;运用自适应尺度小波变换算法提取样本边缘、获取草图信息;将提取后的64维样本图像特征描述符按相应存储方式存人数据库、并利用Best-Bin-Fast算法快速得到最近匹配点;最后经过PROSAC剔除错误匹配后用欧式距离搜索数据库中特征点匹配度最高的图片.

Hand-drawn on the screen in the form of a sketch outline to obtain information and express intention has characteristics of more continent and natural,so the mobile devices of humancomputer interaction with touch method becoming the mainstream.Based on the methods of quantitative analysis the experimental control group,the traditional SURF feature matching algorithm was modified and verified,and the searching efficiency was improved.The optimization method including the application of airspace mean smoothing algorithms to make the image smoother and use adaptive scale wavelet transform algorithm to extract a sample edge and obtain sketch information;then 64-dimensional sample image features was extracted by the corresponding descriptor and stored into a database,and Best-Bin-Fast algorithm was used to quickly get closest match points;finally,the most matching picture on featured points in database was found through PROSAC excluding process.Due to the widely application of parametric design in the production of garment,it not only can improve the algorithm efficiency,but also reduce a lot of repetitive fashion designers unnecessary operations,improve production efficiency.