摘要
近些年来,图像识别技术发展迅速,识别精度越来越高,应用领域也越来越广。但是传统算法在大规模复杂场景的识别效率和准确度不高,尤其当场景中有大型建筑物时,由于建筑物复杂的结构以及户外环境如天气、光照、遮挡等因素,特征点多而杂,分布不均匀,难以获取正确的特征信息。文章应用了AKAZE算法来进行户外场景识别,并构建了基于词汇树的海量图像识别框架。实验证明,该识别框架在户外建筑场景中识别准确度和识别效率较高。
In recent years,image recognition technology has developed rapidly,the recognition accuracy has become higher and higher,and the fields of application have also become wider and wider.However,traditional algorithms have low recognition efficiency and accuracy in large-scale complex scenes,especially when there are large buildings in the scene.Due to the complex structure of buildings and outdoor environments such as weather,light and occlusion,feature points are numerous and miscellaneous,uneven distribution,it is difficult to obtain the correct feature information.In this paper,the AKAZE algorithm is applied to the recognition of outdoor scenes,and a massive image recognition framework based on vocabulary tree is constructed.Experiments show that the recognition framework has high recognition accuracy and recognition efficiency in outdoor architectural scenes.
作者
石力
Shi Li(Xi’an University of Architecture and Technology,Xi’an 710000,China)
出处
《无线互联科技》
2018年第15期101-102,共2页
Wireless Internet Technology
关键词
AKAZE
户外建筑场景识别
词汇树算法
AKAZE
outdoor architectural scene recognition
word tree algorithm
