基于结构光的特征点提取方法研究

Research on Feature Point Extraction Method Based on Structured Light

特征提取是基于视觉定位的关键,只有选择合适的特征及特征提取方法才能确保定位精度。选取结构简单、成本低且易实现的单目视觉定位技术作为研究对象,以屋顶上的结构光点作为观测数据来源。为得到耗时少、提取数目满足定位需求的特征点提取算法,对Moravec、Harris、Harris-Laplace、SUSAN、DoG、FAST等6种特征点提取方法进行比较,选出适合该环境下的特征点提取方法。实验表明,在提取结构光环境下,FAST特征点提取耗时优于其它算法,并且提取数目满足定位所需点数目。

Feature extraction is the key of visual positioning,and only the selection of appropriate features and feature extraction methods can ensure the accuracy of positioning.In this paper,the technology of simple structure,low cost and easy to achieve the single vision positioning is used as the research object,and the structural spot of the roof is used as the source of observation data.In order to get less time consuming,extraction number positioning requirements of feature points extraction algorithm,this paper contrast Moravec,Harris,Harris-Laplace,SUSAN,DoG,FAST six kinds of feature point extraction method in extracting the number of structure light and time consuming,finally choose the feature point extraction method is suitable for the environment.The test proved that,in extracting the structure light environment,FAST feature point extraction time is better than other algorithms and the number of extracted points can be obtained.