摘要
多传感器数据融合作为一种特殊的数据处理手段在图像识别领域得到了较大的重视和发展,本文提出了一种融合多传感器信息的移动图像识别方法.首先通过在智能手机端提取带传感器信息的图像局部特征,增强局部特征的辨别能力;其次改进了随机聚类森林的建立算法,减少了样本图像训练时间;最后使用快速几何一致性校验对匹配结果进行检查,保证算法的识别精度.实验结果表明,本文提出的方法能够快速有效地识别移动图像,并具有较好的鲁棒性,同时与传统的Vocabulary tree方法进行比较,本文方法的识别速度和精度较优,训练代价较低.
Multi-sensor data fusion as a special means of data processing in the field of image recognition has been developed rapidly. This paper presents a novel recognition approach for mobile image to fuse multi-sensor data. Firstly, the local features are extracted by fusing the sensor information to enhance the distinguish ability of them; secondly, the established method of random clustering forest is improved to reduce the training time of sample images; finally, the fast geometric consistency approach is used to check the matching result to ensure the recognition accuracy. Experimental results show that the proposed method can quickly and efficiently recognize the object and has strong robustness. It also has a higher accuraey, faster recognition speed, and less training complexity than the traditional method of vocabulary tree.
出处
《自动化学报》
EI
CSCD
北大核心
2015年第8期1394-1404,共11页
Acta Automatica Sinica
基金
国家高技术研究发展计划(863计划)(2013AA013802)
国家自然科学基金(61370134)
国家重大科技专项(2012ZX03002004)
广东省协同创新与平台环境建设专项(2014B090901024)资助~~
关键词
多传感器数据融合
移动图像识别
随机聚类森林
智能手机
Multi-sensor data fllsion, mobile image recognition, randomized clustering forests, smartphone
