Anomaly detection method for urban river water quality based on background model prediction using three-dimensional fluorescence
Qiuhao Zhang,Jie Yu,Jiegen Shi,Fangjia Xue,Dibo Hou,Guangxin Zhang
DOI: https://doi.org/10.1088/1361-6501/ad5748
IF: 2.398
2024-06-14
Measurement Science and Technology
Abstract:The use of three-dimensional fluorescence spectroscopy for anomaly detection of urban river water quality can protect the water environment from exogenous pollution and organic pollutant toxicity, and has great potential for application. However, current detection methods inadequately address critical application scenarios, such as fluctuations in river water background, detection challenges posed by low concentration pollutants, and the fluorescence peaks overlap between pollutants and background. In this paper, a fluorescence spectrum feature extraction method which is effective for the above scenarios was proposed. The proposed method involves a sequential process. First, the original spectrum undergoes preprocessing using a novel method. Next, an alternating residual tri-linearization technique is applied to establish a predictive model for river water spectrum changes. Subsequently, the background model is utilized for spectrum decomposition and reconstruction of the test sample. This reconstructed spectrum is then used to derive the residual spectrum by comparison with the original. Finally, frequency domain features are extracted from the residual spectrum to enable classification, while the background model undergoes real-time updates. The results demonstrate the effectiveness of the proposed method in the three scenarios mentioned above, leading the potential of proposed method in in-situ anomaly detection for river water environment is also validated.
engineering, multidisciplinary,instruments & instrumentation