基本信息:
- 专利标题： 一种融合深度学习与背景差法的平安城市车流统计方法
- 专利标题（英）：Deep learning and background difference method fused Safe City traffic flow counting method
- 申请号：CN201710801432.6 申请日：2017-09-07
- 公开(公告)号：CN108074244A 公开(公告)日：2018-05-25
- 发明人： 厉紫阳 , 沈徐兰 , 冯卢梦 , 周红晶
- 申请人： 汉鼎宇佑互联网股份有限公司
- 申请人地址： 浙江省杭州市延安路511号元通大厦1119室
- 专利权人： 汉鼎宇佑互联网股份有限公司
- 当前专利权人： 海峡创新互联网股份有限公司
- 当前专利权人地址： 浙江省杭州市延安路511号元通大厦1119室
- 代理机构： 杭州君度专利代理事务所
- 代理人： 朱月芬
- 主分类号： G06T7/11
- IPC分类号： G06T7/11 ; G06T7/194 ; G06T7/246 ; G06K9/62
The invention discloses a deep learning and background difference method fused Safe City traffic flow counting method, and aims at overcoming defects of infrared, ground induction coil and supersonicwave detection methods. A background difference method is used to obtain an object in an image, and the object is trained and classified via deep learning. CNN and GAN networks are used to classify the objects to be identified, a determining axis and an identification area are set dynamically according to a classification result, and vehicles are identified and counted. A higher counting precisionis realized in different environments, the method is adapted to model training under the condition that training samples are not rich, data features can be extracted more accurately, and the model classification accuracy is improved.
公开/授权文献:
- CN108074244B 一种融合深度学习与背景差法的平安城市车流统计方法 公开/授权日：2021-05-25