稀疏性是回归分析和信号处理领域的重要科研课题，具有突出的理论和实际意义。本项目研究稀疏性与轮廓提取问题相结合时所涉及的理论和应用问题，着眼于促进应用背景下稀疏性理论知识的发展，并希望通过引入稀疏性先验知识降低现有轮廓提取方法对人工干预的依赖。具体研究内容包括：（1）利用稀疏表示具有较强特征选择能力的特点，设计轮廓关键点自动筛选方法，从连续曲线轮廓中自动提取出能准确描述整个轮廓的少数关键位置点，这是轮廓建模的关键步骤；（2）通过引入稀疏性约束保证在只拥有部分关键点位置信息时，仍然能计算出完整轮廓在当前字典下的表示系数，从而实现基于部分关键点的整体轮廓重构，有效降低人工标记关键点的工作量；（3）基于方程组的稀疏解唯一性存在理论，设计稀疏表示的松弛范围，使得稀疏性理论知识能够与实际问题相结合；（4）设计描述字典在线更新策略。

Sparsity is an important research topic in regression analysis and signal processing, which has outstanding theoretical and practical values. In this project, we will discuss both theory and application problems occurred during the combination of sparse prior and shape extraction. Through this research, we look forward to promote the development of sparsity theory in practical applications, and reduce the manual interventions in shape extraction algorithms after enrolling sparse priority. Our specific research contents include, (1) design automatic shape landmark selection method that takes advantage of the excellent feature selection ability of sparse representation, which could accurately pick up representative discrete key points from continuous curves, a critical step of shape modeling; (2) through introducing sparse constraints, guarantee the acquisition of sample representation coefficients under description dictionary only given part of landmark pints, and hence realize partial manual shape based shape reconstruction, which could effectively reduce the amount of manual marking; (3) since perfect sparse representation is impossible in real applications, we will analyze and design sparse relaxation range based on the uniqueness theory of sparse linear system of equations to make sparsity real applicable; (4) design online dictionary updating strategy.