摘要
智能手机强大的拍摄功能使得通过手机拍摄屏幕来获取敏感信息的代价越来越低,存在较大风险。在分析现有屏幕防拍摄方法存在的局限性基础上,针对屏幕防手机拍摄的特定环境,提出了一种基于Faster RCNN的识别手机拍照的新方法。首先,分析使用手机拍照的特点,建立了动静结合的Pascal VOC训练集;其次,在TensorFlow的框架下实现手机识别算法;最后,根据不同的影响因素建立训练集,分析各因素在不同的阈值下对识别精准率、准确率和召回率的影响。通过对算法的详细分析和与其他屏幕防窃拍方法进行对比,验证了算法的有效性。
Smartphone has powerful shooting capabilities and makes it cheaper to get sensitive information by taking pictures on personal phone. There are big risks in this matter. Based on the analysis of the limitations of existing screen anti-shooting methods,aiming at the specific environment of screen-proof mobile phone shooting,a new method based on Faster RCNN is proposed to identify mobile phone photos. Firstly,the characteristics of using mobile phone to take pictures are analyzed,and a dynamic and static Pascal VOC training set is established. Secondly,the mobile phone recognition algorithm is realized under the framework of TensorFlow. Finally,the training set is established according to different influencing factors,and the influence of each factor on the recognition accuracy,precision and recall rate under different thresholds is analyzed. The algorithm is analyzed in detail and is compared with other screen anti-theft photography methods. It verifies the effectiveness of the algorithm.
作者
王晓媛
张文涛
韩磊
Wang Xiaoyuan;Zhang Wentao;Han Lei(China Aerospace Academy of Systems Science and Engineering,Beijing 100037 ,China)
出处
《航天控制》
CSCD
北大核心
2019年第2期66-72,共7页
Aerospace Control
