Research On Indoor Positioning Method With The Fusion Of Binocular Vision Sensor And IMU

With the development of mobile internet technology,outdoor positioning technology has gradually matured and has been widely used in various fields such as navigation,aviation,natural disaster prevention,and vehicle navigation.Indoors,with the complexity and diversity of buildings the demand for location services is growing stronger.The existing indoor positioning methods have their own characteristics,and the visual positioning is gradually paid attention to the advantages of being able to work independently in an unknown environment,without having to arrange related equipment in advance,small in size,low power consumption,and rich information acquisition.The INS based positioning method is continuously utilized by the indoor positioning field with high real-time performance,strong anti-interference ability,low cost,and low power consumption.This paper has carried out research and analysis on the following aspects around visual positioning and inertial navigation:(1)The visual positioning algorithm based on binocular camera is studied.The camera imaging model and distortion model are established.The camera is calibrated by Zhang's calibration method to obtain the internal and external parameters and distortion parameters of the camera.The feature points in the image acquired by the binocular vision camera are extracted,and the feature points of the left and right eye images at the same time are matched.The 3D distance calculation is performed while the pose of the carrier is estimated by matching the feature points of the adjacent two frames of images.Then the method of graph optimization and loopback detection in visual positioning is introduced.Finally,the characteristics of the positioning method are briefly described.(2)The inertial navigation system was studied.Firstly,the coordinate system commonly used in inertial navigation,the method of rotation representation between coordinate systems and the conversion method between different rotation representation methods are expounded.Secondly,the method of sensor calibration and initial alignment in inertial navigation system are studied.Then,In the INS,the gyro directly integrates the angle value to increase the error with time.The complementary filter is introduced to correct the gyroscope reading.Finally,the heading attitude,velocity displacement solving method and the characteristics of the positioning method are introduced.(3)The indoor positioning method suitable for the fusion of inertial navigationand binocular vision is studied.Inertial navigation has the characteristics of completely autonomous measurement and stable acquisition of attitude data,which can effectively compensate the loss of visual SLAM in the case of poor feature and improve the stability of system positioning.High-precision visual positioning result can effectively correct the accumulated error existing in the inertial navigation positioning and improve the accuracy of the system positioning.Based on the above characteristics and the premise of running in an unknown environment,the fusion and positioning methods of vision and inertial navigation based on extended Kalman filter are determined,and the corresponding state prediction model and correction model are established.Finally,the experimental verification of the proposed algorithm is carried out.The experimental results show that the angle value obtained by integrating the complementary filter correction is smaller than that obtained by integrating the gyro reading directly.The fusion positioning method based on extended Kalman filter is compared with the binocular vision alone.The positioning method and the inertial navigation method have smaller positioning errors,and the obtained trajectory map is closer to the preset trajectory.