Research On Plastic Mulch Mapping Method Based On Time Series Remote Sensing Data

Plastic film mulching cultivation technology can promote crop development,improve crop yield and improve crop quality.However,the use of plastic mulch leads to the accumulation of microplastics in soil,and microplastics in the soil environment for a long time will affect the soil properties and structure,and then do harm to the growth and development of animals and plants in the soil,change the soil microbial community,and even enter the human body through the food chain and respiratory system,posing a threat to human health.Because of the clear specifications of agricultural plastic mulch,it is an important work to quickly and accurately obtain the spatial distribution of plastic mulch to understand the usage and degradation of plastic mulch,protect agricultural production,protect ecological environment and human health.In this paper,Qiu county of Hebei Province,Tumed Left Banner of Inner Mongolia Autonomous Region and Ningcheng County of Inner Mongolia Autonomous Region are taken as the research areas.Through the construction of time series remote sensing data sets,the spectral characteristics of plastic mulch and non-plastic mulch objects are analyzed,the sensitive remote sensing bands of plastic mulch are selected,and the key period of plastic mulch remote sensing recognition is established normalized plastic mulch index(NPMI)was used to analyze the feasibility of this index for remote sensing recognition of plastic mulch.The results are as follows:(1)Through the analysis of the spectral reflectance of plastic mulch in the study area,it is found that plastic mulch covered area has hyperspectral reflectance.In the blue,green,red,swir1,swir2 bands,the spectral reflectance of the plastic mulch is significantly higher than that of water,woodland,shrub and other ground objects,and in the visible light band,the spectral characteristics of plastic mulch and buildings are relatively close.In general,in red and swir2 bands,the spectral reflectance of plastic mulched farmland is significantly different from that of non-plastic mulched crops,plastic greenhouses,water bodies,buildings,woodlands and shrubs,and other ground objects,which indicates that red and swir2 bands are more sensitive to plastic mulch,and can be used as bands to construct normalized plastic mulch index.(2)Based on the time series remote sensing data,using the change of spectral reflectance of plastic mulch in crop growing season,that is,in the stable period of plastic film mulching(late April to early May),the spectral differences between plastic film mulched farmland and non-plastic film mulched crops,plastic greenhouses,water,forest land and shrubs,and other ground objects are large,and the reflectance of plastic film reached the highest value in May,and then gradually decreased,in August Reach the lowest value.Therefore,according to the periodic variation of plastic mulch reflectance in the selected study area,May and August are the key periods to construct the normalized plastic mulch index by using the seasonal index method.(3)Based on the time series remote sensing data,the normalized plastic mulch index(NPMI_Red,NPMI_SWIR2)was constructed to expand the spectral difference between plastic mulch and non-plastic mulch land cover types.The results showed that the normalized plastic mulch index value of plastic mulch covered area was higher than that of non-plastic mulch covered area,which indicated that the index could achieve the purpose of separating plastic mulch and non-plastic mulch land cover.Through the analysis of the separability index-M,it is concluded that the NPMI_Red index is better for the separation of film mulched and non-film mulched ground objects.The threshold of Qiuxian,Ningcheng and Tumed is 0.41,0.47 and 0.44 respectively.Due to the uncertainty of the OSTU threshold method,there is a difference between the threshold determined by OSTU and the optimal threshold(the difference is 0.27 in Qiu County,0.03 in Ningcheng,and 0.13 in Tumed),Affected the accuracy of plastic mulch recognition.Through the analysis and research of the plastic mulch recognition results based on the random forest(RF)and support vector machine(SVM)classification methods,it was found that the plastic mulch recognition accuracy based on the random forest classification method was the highest in Qiuxian,Ningcheng and Tumed research areas.The classification accuracy reached 96.24%,80.27%,and 82.60%,and the Kappa coefficient reached 0.92,0.60,and 0.59.Secondly,the classification method based on support vector machine,the overall classification accuracy was 95.31%,75.20%,and 74.95%,respectively,and the Kappa coefficient was respectively 0.90,0.51,and 0.46.Experiments show that it is feasible to use the normalized plastic mulch index to identify the plastic mulch,and it can effectively carry out large-scale plastic mulch recognition.