【作者】 王立； 吴羽溪； 王小艺； 刘载文；

【Author】 WANG Li;WU Yuxi;WANG Xiaoyi;LIU Zaiwen;Beijing Key Laboratory of Big Data Technology for Food Safety, School of Artificial Intelligence, Beijing Technology& Business University;

【机构】 北京工商大学人工智能学院北京市食品安全大数据技术重点实验室；

【摘要】 化工生产过程中排放过量的氮和磷元素导致水体富营养化及藻类水华灾害频发,而现有的藻类水华分析存在两方面问题:一方面仅考虑氮或磷元素对藻类生长过程影响,忽略水体底物中的氮磷反馈机制,没有反映氮和磷元素在藻类水华形成过程中的完整循环过程;另一方面仅基于现场监测点数据研究,缺乏对整体水域的全面分析,而遥感图像能够反映整体水域变化,但传统分析方法难以有效处理海量的遥感数据。本研究以富营养化状态和叶绿素a浓度作为藻类水华的表征指标,以遥感图像及水体中的总氮和总磷为主要研究对象,提出一种基于遥感图像3D-CNN及氮磷循环的水华形成过程分析新方法。首先,基于3D-CNN对遥感图像进行特征提取,并采用细菌觅食算法优化网络结构,预测水体富营养化等级。其次,考虑水华形成过程中氮循环、磷循环等化学过程,根据"氮-磷-藻"之间的耦合关系及底物反馈机制,建立三维生化动力学时变参数模型,确定水华暴发程度及临界条件,并融合遥感图像提取的特征信息建立ENN模型,预测水华暴发的时间及程度。本研究选用由MODIS卫星获取的太湖流域遥感图像及水域中的总氮和总磷等水质监测数据。仿真结果表明,基于遥感图像3D-CNN结合氮磷循环的分析方法在富营养化状态和水华暴发预测方面均取得良好效果。更多还原

【Abstract】 Excessive nitrogen and phosphorus emissions in the chemical production process lead to eutrophication and frequent algal blooms. However, most of the existing algal blooms analysis is based on the monitoring data of the water area, which lacks the comprehensive analysis of the whole water area. At the same time, although the remote sensing image can reflect the changes of the whole water area, the traditional analysis method is difficult to effectively process the massive remote sensing data. In this study, the eutrophication state and chlorophyll-a concentration were taken as the characterization indexes of algal blooms, and the remote sensing images and the total nitrogen and phosphorus in the water were taken as the main research data,then a new method based on remote sensing image 3 d-cnn and nitrogen and phosphorus cycle is proposed. Firstly, the remote sensing image features are extracted based on 3 D-CNN, and the network structure is optimized by bacterial foraging algorithm to predict the water eutrophication level. Secondly, considering the chemical processes such as nitrogen and phosphorus cycle in the process of bloom formation, according to the coupling relationship between ‘nitrogen-phosphorus-algae’ and substrate feedback mechanism, a three-dimensional biochemical kinetic time-varying parameter model was established to determine the outbreak degree and critical conditions of the bloom. Combined with the characteristic information extracted from remote sensing images,the ENN model was established to predict the time and extent of the bloom. In this study, the remote sensing images of Taihu Lake Basin obtained by MODIS satellite and the water quality monitoring data such as total nitrogen and total phosphorus were selected. The simulation results show that the analysis method based on remote sensing image 3 D-CNN combined with nitrogen and phosphorus cycle achieved good results in eutrophication status and water bloom outbreak prediction.更多还原