Research On Detection Of Soil Oil Pollution In Oil Exploration Areas Of The Yellow River Delta Based On RS

Oil is one of the main human energies, following by the expansion of the development and exploitation, the pollution on soil and environmental is growing more serious. Oil into the soil, not only will destroy the soil and affect plant growth, but substances can also pose threats of many levels to human and its living environment by the groundwater pollution and the transfer of food chain contamination. Effective management and remediation of oil pollution is based on the premise that the contaminated area and the degree of pollution are determined. The development of remote sensing technology provides new, more efficient and convenient means for the detection of oil pollution.This paper used measured contaminated soils spectral curve and ASTER remote sensing image combined with the spectral characteristics of petroleum-series of substance and based on remote sensing technology to study the remote sensing extraction of spectrum of oil content of soil inversion and abnormal information. Firstly, we collected soil samples in experimental petroleum contaminated areas, obtained sample points location and petroleum content data, and then obtained soil sample hyperspectral data by using AvaField geophysics spectrometer in laboratory room. Secondly, for soil sample spectra data and petroleum content data, we used the correlation analysis and multiple stepwise analysis to ascertain spectrum characteristic parameters of soil petroleum substances and establish multivariate linear statistical model and the BP neural network model through a variety of spectral transform analysis, and inspected models precision, compared and analyzed these two kinds of models. The result showed that BP neural network model was more feasible and had more advantages in the simulation and forecast of oil content in the soil; in the next place, it effectively improved the accuracy of classification for using FLAASH model to finish ASTER remote sensing image atmosphere correction, denoising and compressing image spectrum data utilizing MNF and PCA after having obtained surface features'spectral reflectance, extracting texture feature by using gray level co-occurrence matrix, and using SVM method to classify and identify the image synthetic by spectral features and texture features. Finally, we finished the sorted bare soil's petroleum content anomaly information's extraction; fabricated chart of petroleum content anomaly information's extraction, and verified extraction results by measured data.The experiments proved that BP neural network could be better to simulate relationship between the petroleum content in soil and their spectral data, and had advantages in building soil petroleum content spectral inversion model. It could improve classification accuracy by using support vector classification with the MNF transformation and the texture characteristics. Petroleum content anomaly information extracted by remote sensing image ASTER was practical, which proved that it was feasible to obtain petroleum storage anomaly information with the help of remote sensing tools, but the precision needed improving.