| With the development of remote sensing technology and the progress of deep learning in image processing,the method of using deep learning to recognize remote sensing image has been widely used in various fields,such as agriculture,forestry,environment and so on.Tea is one of the most important economic crops in China’s agriculture.How to obtain the planting area of tea accurately and timely is of great significance for the government to master the production of tea and formulate relevant control policies.The non-tea planting area in the remote sensing image is much larger than the tea planting area.In order to take into account the efficiency and accuracy of the algorithm,this paper divides the entire recognition process into two stages: rough division of tea planting area and segmentation of tea planting area.The rough division of tea planting areas is mainly for the purpose of obtaining a wide range of tea planting areas and removing a large number of non-tea planting areas,so as to reduce the amount of data of subsequent tea area segmentation.The rough segmentation model firstly divides the large-scale remote sensing image into small regions,and marks the small images into two categories: tea planting area and non-tea planting area.By establishing a classification algorithm based on convolutional neural network and adopting transfer learning,a large number of non-tea planting areas are removed to obtain the rough division of tea planting area.In order to obtain a more accurate tea planting area,the isolated point analysis algorithm was used to eliminate the misclassified tea planting area in a small area,which improved the accuracy of the algorithm.Based on the roughly divided tea planting area,a tea planting area segmentation model is established by constructing the image segmentation training data set and adopting the full convolution neural network.The influence of the full convolution network model of different feature extractors on the segmentation effect is discussed,and an effective segmentation model is obtained,which provides a basis for the extraction of tea planting area. |
