Agricultural facility remote sensing extraction method and device based on fine classification

Abstract

The invention relates to an agricultural facility remote sensing extraction method and device based on fine classification. The agricultural facility remote sensing extraction method introduces a deeplearning technology into agricultural facility remote sensing extraction, realizes full-automatic acquisition of agricultural facility contour boundaries, the number and the area, avoids the complexfeature selection process in a traditional pixel classification method and an object-oriented classification method, remarkably improves the agricultural facility extraction precision, and saves manpower and time. The agricultural facility remote sensing extraction method also subdivides agricultural facilities into common single plastic agricultural facilities, multi-span agricultural facilities,multi-span or single agricultural facilities covered with sunshade nets, greenhouse agricultural facilities, agricultural facilities without thin film coverage and equipment rooms. The agricultural facility remote sensing extraction method performs training and extracting on the agricultural facilities in different scenes, and fuses the results of the different types of agricultural facilities inthe final result.

Description

technical field [0001] The application belongs to the field of remote sensing extraction of agricultural facilities, and in particular relates to a method and device for remote sensing extraction of agricultural facilities based on fine classification. Background technique [0002] In recent years, with the rapid development of agricultural modernization and the gradual maturity of agricultural facilities and technologies, the area of ​​agricultural facilities in my country has rapidly expanded, and China has become a large agricultural facility in the world. Agricultural facilities have become an important land cover type. The extraction of traditional agricultural facilities is mainly based on artificial features and machine learning classifiers. There are problems such as complex feature selection, low recognition accuracy, long processing time, and inability to accurately obtain the boundaries and quantities of greenhouses. The acquisition of vector boundaries of agricu...

AI Technical Summary

Problems solved by technology

This method is time-consuming and laborious, and requires digitizers to have rich geoscience knowledge and experience in image interpretation, which is prone to errors

Regardless of the combination of spectrum and texture or the object-oriented method, the accuracy of automatic extraction of agricultural facility greenhouses is not high, mainly because of the distribution of various types of agricultural facilities in the area, high density, and pixel confusion between translucent films and ground objects planted inside. , a variety of materials, etc., which can specifically include the following characteristics: different colors of transparent roofs (blue, baise, etc.), different roof materials (plastic film, glass, fiberglass), large thickness differences, different light reflection and absorption laws, and greenhouses. There are various kinds of crops (vegetables, seedlings, grapes, etc.) and different greenhouse coverings (insulation quilt, some are completely open, some are covered with sunshade nets, insect screens, grass curtains, etc.), which makes the remote sensing images of agricultural facilities in the spectrum There are great differences in information, and these characteristics increase the difficulty of identifying the area and quantity of greenhouses

Images