Analysis Of Land Use Change And Driving Forces In Northern Henan Plain Based On Semi-Supervised Machine Learning

Land use is an important research issue in global environmental and ecological change.Dynamic monitoring of land use change provides a certain theoretical basis for land use decision making.Land use and land cover(LULC)classification using remote sensing images is an effective method,which plays an important role in many environmental modeling and land use planning.In order to solve the problems of high cost of sample label acquisition and serious misclassification and omission of classification results in previous land use classification algorithms.In this paper,a semisupervised classification framework based on PSO probabilistic neural network is proposed to exploit and utilize as much information as possible to achieve land use classification of remote sensing images.The feasibility of this study was confirmed in the northern Henan plain.Firstly,particle swarm optimization algorithm was used to optimize the parameters of classifier,and the accuracy of probabilistic neural network classifier was improved.Shannon entropy was used to select samples with high confidence to expand the initial training sample set,and a large number of unlabeled samples were extended to the training sample set,and the number of initial labeled samples was reduced.The spatial-temporal changes and driving forces of land use in north Henan plain were analyzed by using remote sensing images in 1996,2001,2008,2014 and 2020.The results provide a certain reference for further rational development of land resources in north Henan plain,and also provide countermeasures for sustainable development of land resources.The main research work of this paper is as follows:(1)This paper proposes a new land use classification method of remote sensing images.Based on PSO probabilistic neural network and semi-supervised machine learning,a land use classification framework based on semi-supervised machine learning was proposed.The aim is to solve the problems of low classification accuracy,high classification cost and difficult model construction in machine learning algorithm.In this paper,the northern Henan plain is selected as the research area,and the research is divided into six categories: cultivated land,construction land,forest land,grassland,water body and other land.After the classification system was determined,the proposed algorithm was compared with random forest method,maximum likelihood method and probabilistic neural network algorithm for land use classification,and two accuracy evaluation indexes of overall accuracy and Kappa coefficient were selected to determine the classification accuracy.The results show that the classification accuracy of the proposed algorithm is more than 3% higher than the other three algorithms,and the Kappa coefficients are all above 0.8.Therefore,the new algorithm proposed in this paper can be effectively applied to land use/cover classification of remote sensing images.(2)Land use change: Based on the new algorithm proposed in this paper,the land use classification results of five periods in north Henan plain were obtained.Based on geographic information technology platform,the dynamic attitude of land use and land use transfer matrix of north Henan plain were analyzed,and the spatio-temporal change characteristics were explored.The results showed that the construction land in northern Henan plain increased from 1996 to 2020,and the cultivated land area decreased first and then increased a little.The changes of forest land and water area were not obvious,and other land areas were less year by year.Among them,the fastest increase is construction land,the fastest decrease is other land.In addition,the land use transfer matrix is used to reveal the law of land transfer in north Henan plain.The results show that the main direction of grassland transfer is farmland and construction land.Farmland is mainly transferred to construction land and other land;The main circulation of woodland is grassland;Part of water body is developed as construction land and part as arable land;Other land areas will be converted into arable land,grassland and construction land.(3)Analysis of driving forces of land use change: Based on the social and economic statistical data of each region in north Henan Plain,the evaluation system of driving forces of land use change in north Henan plain was constructed,and the driving forces were analyzed by principal component analysis.It can be found that the increase of water body and population is the main driving force of land use/cover change in north Henan plain.Of economic factors,the main impact factor is the growth of social economy,especially the growth of the second and third industry,leading to social development to the increasing demand of construction land and other resources,and social factors of land use/cover change in the plains of north secondary position,while the natural factors also can cause the land use change,but its effect is smaller.It provides technical support for the formulation and reform of land policy in north Henan Plain.