Research On Key Technologies Of Aerial Remote Sensing System Based Small UAV

Since the 80 s of last century,because the popularization of computer technology promotes the development of remote sensing technology,the application scope of remote sensing technology is expanded from military field to civil.However,with the acceleration of the process of national information construction,the traditional remote sensing technology exposes many shortcomings,such as low real-time,high operating costs,easy to be affected by the weather and so on.With the rapid development of sensor technology and automatic control technology,the technical threshold and the development cost of UAV have been reduced.The UAV system equipped with remote sensing imaging equipment will be able to make up for the shortcomings of traditional remote sensing systems,reduced system operating costs,increasing real-time remote sensing data,improving operational flexibility,and greatly promoting the development of new generation of remote sensing technology.This paper studies two key technologies: fixed-wing UAV remote sensing platform and remote sensing image processing.The main research contents are as follows:(1)The paper summarizes the background,development and research status of remote sensing technology,and introduces the software and hardware system of UAV flight control system and the related algorithms of hyperspectral remote sensing image processing.(2)A key technology of UAV remote sensing platform is designed and implemented on the embedded system-attitude reference system.The attitude reference system designed in this paper is the premise of the UAV's autonomous and steady flight,which provides important parameters for UAV control and navigation,the correction and processing of the remote sensing image.The quality of the attitude reference system is directly related to the flight accuracy and the quality of the remote sensing image.In this paper,the design of UAV flight control system uses MEMS inertial sensors,which is costeffective but low measurement accuracy.If the measurement error is not compensated,it will result in the divergence of attitude calculation results.To solve the above problems,three-axis accelerometers and magnetometers are used to compensate the drift error of the three-axis gyro by multi sensor fusion.(3)The paper presents a spatial-spectral feature extraction method for hyperspectral remote sensing data based on deep learning.Because hyperspectral data has the characteristics of nonlinear,large data redundancy,high spatial correlation and so on,it is very easy to cause "dimension disaster" in the classification of hyperspectral images without feature extraction.To get the mapping function from hyperspectral data to nonlinear feature and study the nonlinear feature of high dimensional data effectively,a deep model is constructed-stacked auto encoder machine,through the classical deep learning model-autoencoder machine.Finally,spatial information is introduced into the nonlinear deep feature by spatial-spectral combination algorithm,which further improves the classification accuracy.In this paper,the remote sensing platform attitude reference system and the remote sensing image processing algorithms for small fixed-wing UAV remote sensing technology are studied.The direction cosine matrix method and multi-sensor fusion are used to realize the attitude reference system;a spatial-spectral feature extraction method based on deep learning model for hyperspectral data is proposed,which extracts the nonlinear features and spatial information effectively.Finally,the effectiveness of the above design is verified by comparison experiments.