Research On Tidal Cycle Monitoring Of Surface Suspended Sediment Concentration In The North Port Of The Yangtze Estuary Based On UAV Hyperspectral

Suspended Sediment Concentration(SSC)not only directly affects the optical properties such as transparency and turbidity of the estuary estuary,but also affects the change of estuaries and soil and water conservation in coastal areas.Studying the spatial distribution and change of SSC in the estuary is of great significance for maintaining the health of coastal areas and marine ecosystems and understanding the topographic and geomorphic changes of the estuary.With the development of remote sensing technology,the quantitative inversion of surface suspended sediment concentration(SSSC)by using remote sensing reflectance of water has become an important method of suspended sediment monitoring.At present,most of the inversion research is based on satellite multispectral data,which has some problems,such as low spatial resolution,incomplete spectral information,unable to carry out flexible monitoring and so on.In recent years,UAV remote sensing has been applied to water quality monitoring,but at present,the research on SSSC monitoring based on UAV technology is still less,especially for the estuarine area with complex hydrodynamic environment,the UAV hyperspectral inversion model is not perfect,and there is a lack of SSSC change research in tidal cycle.In view of the above problems,this study selects the north port of the Yangtze River Estuary as the experimental area,and proposes to use the UAV equipped with hyperspectral sensor to monitor the change of SSSC in the tidal cycle.By analyzing the spectral curve characteristics of sediment laden water,the sensitive band of SSSC is determined.On this basis,the SSSC hyperspectral inversion model is studied,and the model is applied to the SSSC inversion of UAV image to analyze the variation law of SSSC under different tide conditions in the north port of the Yangtze Estuary.The main research work and conclusions are as follows:(1)Aiming at the influence of suspended sediment concentration on the optical properties of the water body in the Yangtze River Estuary,the radiation transmission principle of light in the water body is expounded.The surface sediments with typical particle size near the south mouth of the Yangtze River estuary are collected,and the quantitative study on the spectral characteristics of SSSC gradient change in the water body of the Yangtze River estuary is carried out.The results show that with the increase of SSSC,the spectral reflectance curve of water body has similar morphology,but there are also some differences.When SSSC<36.59mg/l,there is only one reflection peak at560～580nm in the spectral curve;When SSSC>52.69mg/l,the first reflection peak of the spectral curve is 570～710nm,and the second reflection peak appears at 780～820nm in the near infrared band.There is an obvious"red shift"phenomenon in the first peak interval of the spectral curve,and the linear fitting R~2 between the"red shift"change of wavelength and SSSC reaches 0.91,indicating that there is a strong linear correlation between them.In the spectral range of 400～900nm,the band in the range of 740～860nm has strong response ability to the change of SSSC and is relatively stable.Therefore,740～860nm is selected as the sensitive band interval of SSSC change,which provides a theoretical basis for the next study of SSSC hyperspectral inversion model.(2)Aiming at the problem of data redundancy caused by the large number of hyperspectral bands and strong linear correlation between adjacent bands,this study introduces a hyperspectral band dimension reduction method based on principal component analysis.A small number of uncorrelated principal component variables are used to replace a large number of original hyperspectral band data,and the original spectral information is retained as much as possible while reducing the model input parameters,so as to simplify the hyperspectral data set.This paper analyzes the dimensionality reduction results of hyperspectral bands of two different data sets in the north port of the Yangtze River Estuary in wet season and dry season.It is found that for different data sets,the first three principal components have been able to map more than90%of the information of the original spectral data,and the first 10 principal components have been able to map more than 99%of the information of the original spectral data.It shows that using 3～10 principal component variables can replace 45 bands of the original spectral data.(3)The PCA-BP Neural network model is established by selecting 3～10 principal component variables and using the data sets of two different periods of wet season and dry season,and compared with the traditional linear regression model.The results show that the prediction accuracy of PCA-BP Neural network model established in different periods is better than that of linear regression model.For the wet season,when six principal component variables are selected,the fitting R~2 of PCA-BP inversion model reaches 0.87,RMSE is 18.45mg/l and MRE is 9.41%.In dry season,when five principal component variables are selected,the fitting R~2 of PCA-BP model is 0.93,RMSE is4.21mg/l and MRE is 8.75%.(4)Select the Beigang reach of the Yangtze River estuary where the suspended sediment concentration is obviously affected by the tide,and carry out the monitoring research of SSSC tidal cycle change based on UAV hyperspectral sensor.The PCA-BP Neural Network Model Based on the principal component analysis method is used for SSSC inversion,and the SSSC change monitoring under different tide conditions of flood season,dry season,flood tide,flood rest and falling rest is completed.The experimental results show that the average value of SSSC in wet season is higher than that in flood tide and higher than that in fall.With the increase of offshore distance,the SSSC in flood tide changes sharply,and the SSSC in flood and fall periods fluctuates less;In dry season,the average value of SSSC is higher than that of fallow,and SSSC also increases with the rise of tide level.SSSC in different periods first decreases and then tends to be stable with the increase of offshore distance.Comparing the inversion results in different periods,it is found that the SSSC in wet season is higher than that in dry season.It is verified that the inversion results are consistent with the measured results,which shows that the SSSC tidal cycle monitoring method proposed in this study based on UAV is reliable.