Research On UAV Navigation Technology Combining Vision And Inertial Navigation

With the continuous progress and development of science and technology,UAV is widely used in various fields,and UAV navigation technology is becoming the focus of researchers.Based on the research of UAV navigation technology,this dissertation aims to improve the accuracy of UAV navigation technology,and takes the loosely-coupled combinatorial structure of inertial navigation assisted by computer-vision as the framework.Some key navigation technologies,such as the principle of strapdown inertial navigation system(SINS),MEMS gyroscope noise analysis and processing,computer vision positioning principle and integrated navigation information fusion method,are studied in this dissertation.Its main content can be summarized as:1、This dissertation studies the transformation of the coordinate system,the attitude and position update,and the principle of SINS.2、In order to improve the measurement accuracy of MEMS gyroscope signal,the noise characteristics of MEMS gyroscope are studied by Allan variance analysis.And a denoising method of MEMS gyroscope combining Kalman and wavelet is proposed in this dissertation.Firstly,the ARMA model is established.According to the model,an adaptive anti-field Kalman filter is designed.The filter can real-timely detect the interference data in the observation data according to the information,and the influence of the outliers on the filter are avoided by modifying the gain and the predicted value.At last,a new wavelet denoising method is used to denoise the filtered signal.In this wavelet denoising method,the wavelet decomposition level is determined,and then it uses Wavelet analysis to process the low and high frequency components simultaneously.Experimental results show that this method can effectively improve the accuracy of the sensor and reduce the error.3、This dissertation studies the imaging principle and calibration of the camera as well as the positioning principle of computer vision.Aiming at the problems of low image matching rate and uneven feature distribution in computer vision positioning,an adaptive ORB feature extraction method based on image segmentation is proposed.The method adopts adaptive image segmentation,and controls the total number of features to be extracted from each sub-image by area ratio.What’s more,Fast14-24 corner detection is introduced to avoid the extraction of false corner points.The effectiveness of the method is proved by experiments.4、This dissertation studies the information fusion method of computer-visual/inertial integrated navigation with the loose coupling structure.An extended Kalman filter is used to fuse visual/inertial navigation information.This method takes the errors calculated by inertial navigation as the state,and the difference between visual positioning information and inertial navigation positioning information as the observe.An optimal estimation of inertial navigation parameter error is obtained by a filter.The fusion method is used to correct the inertial navigation errors to improve the accuracy of UAV navigation.