| With the accelerating process of urbanization and the continuous optimization of agricultural structure,cultivated land is increasingly threatened by building occupation,agricultural modernization,returning farmland,production and life and other factors.Cultivated land in many areas of China tends to be non-agricultural and non-grain.In the process of poverty alleviation in Dongxiang Autonomous County,the acceleration of infrastructure construction eradication has a negative impact on the local cultivated land resources.Studying the spatial characteristics and change trend of non-agricultural and non-grain cultivated land in this area can provide a theoretical basis for the optimal regulation of local non-agricultural and non-grain cultivated land.Taking Dongxiang Autonomous County as the study area,this paper carries out relevant research work from three aspects:range extraction,feature recognition and prediction analysis.Firstly,explored the extraction method of non-agricultural and non-grain cultivated land based on remote sensing images,so as to realize the extraction of four periods of non-agricultural and non-grain cultivated land from 2015 to 2019 in the study area;secondly,the spatial characteristics of non-agricultural and non-grain cultivated land in Dongxiang Autonomous County are studied from four aspects:land type composition,slope distribution,landscape pattern and spatial adjacency;finally,the change trend of non-agricultural and n on-grain cultivated land in Dongxiang Autonomous County from 2020 to 2024 is predicted,and the corresponding optimal regulation scheme is put forward.The main conclusions are as follows:(1)Research on the extraction method of non-agricultural and non grain cultivated land based on remote sensing images.①Extraction of non-agricultural cultivated land.The object-oriented rule extraction method was used to analyze the spectral characteristics,vegetation index characteristics and geometric characteristics of the samples.The extraction rules are constructed for different land types to realize the classification.The classification results of five periods of remote sensing images are obtained and the accuracy is evaluated.The superposition analysis is used to obtain four periods of non-agricultural cultivated land.The results show that the overall accuracy of classification based on object-oriented rule extraction method is higher than 82%,and the Kappa coefficient is higher than 0.78.The results of non-agricultural cultivated land can meet the needs of this study;②extraction of non-grain cultivated land.Combined with the recursive feature elimination method based on cross validation and the random forest classification method,the cultivated land planted with main grain crops in the study area is extracted.By calculating the sum of the decision coefficients of the feature combination,it is concluded that NDVI and EVI are the optimal feature combination.The random forest classification method is used to obtain the cultivated land planted with grain crops in the study area in four stages,and the accuracy is evaluated to further obtain the cultivated land without grain in four stages.The results show that:combining the recursive feature elimination method based on cross validation and random forest classification method to extract the cultivated land planted with main grain crops in the study area,the overall accuracy is higher than 79%,and the Kappa coefficient is higher than 0.76.The results of non-grain cultivated land can meet the needs of this study.(2)This paper explores the spatial characteristics of non-agricultural and non-grain cultivated land in Dongxiang Autonomous County from four aspects:land type composition,slope distribution,landscape pattern and spatial adjacency.The results show that:①the characteristics of land type composition:the non-agricultural cultivated land in Dongxiang Autonomous County is mainly grassland type,and the non-grain cultivated land is mainly cultivated land planted with non-grain crops;② slope distribution characteristics:the non-agricultural cultivated land in Dongxiang Autonomous County is concentrated in the slope range of 6°~25°,and the non-grain cultivated land is mainly concentrated in the slope range greater than 25°;③ landscape pattern characteristics:the patch scale and patch fragmentation degree of each non-agricultural cultivated land type in Dongxiang Autonomous County show an increasing trend,the scale of non-grain cultivated land changes slightly,the non-grain cultivated land type of non cultivated crops develops towards the direction of small patch fragmentation degree,and the non-grain cultivated land type of cultivated non-grain crops develops towards the direction of large patch fragmentation degree;④spatial adjacency characteristics:the natural suitability of adjusting non-agricultural cultivated land to non cultivated cultivated land in Dongxiang Autonomous County is higher.Among the types of non-grain cultivated land,the natural suitability of adjusting non cultivated cultivated land to grassland is higher,and the natural suitability of adjusting non-grain cultivated land to non cultivated cultivated land is higher.(3)Based on the prediction model of land use pattern,this paper explores the change trend of non-agricultural and non-grain cultivated land in Dongxiang Autonomous County from 2020 to 2024,and puts forward corresponding optimization and regulation suggestions in combination with its spatial characteristics.The results show that from 2020 to 2024,the non-agricultural transfer of cultivated land in Dongxiang Autonomous County has been restrained to a certain extent,while the non-grain transfer of cultivated land is still in the stage of being promoted.For the optimal regulation of non-agricultural cultivated land in Dongxiang Autonomous County,first is to encourage farmers to use cultivated land correctly from the policy,and second is to guide the transformation of non-agricultural cultivated land to non cultivated land through manual intervention;for the optimal regulation of non-grain cultivated land in Dongxiang Autonomous County,the first is to encourage farmers to plant crops with the same planting attributes as their adj acent cultivated land,the second is to reduce the cost of grain planting and reduce the abandonment rate of cultivated land by improving the mechanization of grain production,and the third is to reasonably plan and layout the cultivated land for planting non-grain crops and centralize the scattered cultivated land for planting non-grain crops. |