针对基于电化学传感器的检测方法存在密度空间分布低、时间延迟等缺点,设计了一种基于图像的实时智能检测方法。模型通过摄像设备来获取实时的数据,融合了信息丰度测量和宽深度学习机制(IAWD)来对检测目标进行智能建模。模型在新提出的DS变换空间中提取两类特征来度量给定图像的IA。为了同时具有记忆和泛化的优势,设计了一种新的宽深度神经网络来学习上述提取的特征与检测目标之间的非线性映射。此外,为了解决当前普遍采用的深度卷积模型在特征提取以及特征利用上的缺陷,实验室设计了自组织多通道深度神经网络,该神经网络的基本结构为由固定模块和嵌套模块组成的分形模块,同时充分利用卷积网络的跳连操作,加强特征提取和融合利用能力。智能检测模型的构建流程如下图所示:
代表性成果:
[1] Gu Ke, Liu Hongyan, Liu Jing, Yu Xiaofeng, Shi Ting, Qiao Junfei. Air Pollution Prediction in Mass Rallies With a New Temporally-Weighted Sample-Based Multitask Learner. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-15.
[2] Gu Ke, Liu Hongyan, Xia Zhifang, Qiao Junfei, Lin Weisi, Daniel Thalmann. PM2.5 Monitoring: Use Information Abundance Measurement and Wide and Deep Learning. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(10): 4278-4290.
[3] Gu Ke, Liu Jing, Shi Shuang, Xie Shuangyi, Shi Ting, Qiao Junfei. Self-Organizing Multichannel Deep Learning System for River Turbidity Monitoring. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-13.