Ghazaleh Panahandeh

Vision-Aided Inertial Navigation Based on Ground Plane Feature Detection

In this paper, a motion estimation approach is introduced for a vision-aided inertial navigation system. The system consists of a ground-facing monocular camera mounted on an inertial measurement unit (IMU) to form an IMU-camera sensor fusion system. The motion estimation procedure fuses inertial data from the IMU and planar features on the ground captured by the camera. The main contribution of this paper is a novel closed-form measurement model based on the image data and IMU output signals. In contrast to existing methods, our algorithm is independent of the underlying vision algorithm for image motion estimation such as optical flow algorithms for camera motion estimation. The algorithm has been implemented using an unscented Kalman filter, which propagates the current and the last state of the system updated in the previous measurement instant. The validity of the proposed navigation method is evaluated both by simulation studies and by real experiments.

Calibration of the accelerometer triad of an inertial measurement unit, maximum likelihood estimation and Cramér-Rao bound

In this paper, a simple method to calibrate the accelerometer cluster of an inertial measurement unit (IMU) is proposed. The method does not rely on using a mechanical calibration platform that rotates the IMU into different precisely controlled orientations. Although the IMU is rotated into different orientations, these orientations do not need to be known. Assuming that the IMU is stationary at each orientation, the norm of the input is considered equal to the gravity acceleration. As the orientations of the IMU are unknown, the calibration of the accelerometer cluster is stated as a blind system identification problem where only the norm of the input to the system is known. Under the assumption that the sensor noises have a white Gaussian distribution the system identification problem is solved using the maximum likelihood estimation method. The accuracy of the proposed calibration method is compared with the Cramér-Rao bound for the considered calibration problem.

Ground plane feature detection in mobile vision-aided inertial navigation

In this paper, a method for determining ground plane features in a sequence of images captured by a mobile camera is presented. The hardware of the mobile system consists of a monocular camera that is mounted on an inertial measurement unit (IMU). An image processing procedure is proposed, first to extract image features and match them across consecutive image frames, and second to detect the ground plane features using a two-step algorithm. In the first step, the planar homography of the ground plane is constructed using an IMU-camera motion estimation approach. The obtained homography constraints are used to detect the most likely ground features in the sequence of images. To reject the remaining outliers, as the second step, a new plane normal vector computation approach is proposed. To obtain the normal vector of the ground plane, only three pairs of corresponding features are used for a general camera transformation. The normal-based computation approach generalizes the existing methods that are developed for specific camera transformations. Experimental results on real data validate the reliability of the proposed method.

Vision-Aided Inertial Navigation Using Planar Terrain Features

The idea is to implement a vision-aided inertial navigation system (INS) for estimating inertial measurement unit (IMU)-camera ego-motion. The system consists of a ground facing monocular camera mounted on an IMU that is observing ground plane feature points. The motion estimation procedure is through tracking detected corresponding feature points between two successive image frames. The main contribution of this paper is a novel closed-form measurement model based on the image data and IMU output signals. In contrast to existing methods, our algorithm is independent of the underlying vision algorithm such as image motion estimation or optical flow algorithms for camera motion estimation. Additionally, unlike the visual-SLAM based methods, our approach is not based on data association. The algorithm has been implemented using an Extended Kalman filter (EKF), which propagates the current and the last state of the system updated in the previous measurement state. Simulation results show that the introduced method is persistent to the level of the noise and works well even with few numbers of features.

Image moments for higher-level feature based navigation

This paper presents a novel vision-based localization and mapping algorithm using image moments of region features. The environment is represented using regions, such as planes and/or 3D objects instead of only a dense set of feature points. The regions can be uniquely defined using a small number of parameters; e.g., a plane can be completely characterized by normal vector and distance to a local coordinate frame attached to the plane. The variation of image moments of the regions in successive images can be related to the parameters of the regions. Instead of tracking a large number of feature points, variations of image moments of regions can be computed by tracking the segmented regions or a few feature points on the objects in successive images. A map represented by regions can be characterized using a minimal set of parameters. The problem is formulated as a nonlinear filtering problem. A new discrete-time nonlinear filter based on the state-dependent coefficient (SDC) form of nonlinear functions is presented. It is shown via Monte-Carlo simulations that the new nonlinear filter is more accurate and consistent than EKF by evaluating the root-mean squared error (RMSE) and normalized estimation error squared (NEES).

Exploiting ground plane constraints for visual-inertial navigation

In this paper, an ego-motion estimation approach is introduced that fuses visual and inertial information, using a monocular camera and an inertial measurement unit. The system maintains a set of feature points that are observed on the ground plane. Based on matched feature points between the current and previous images, a novel measurement model is introduced that imposes visual constraints on the inertial navigation system to perform 6 DoF motion estimation. Furthermore, feature points are used to impose epipolar constraints on the estimated motion between current and past images. Pose estimation is formulated implicitly in a state-space framework and is performed by a Sigma-Point Kalman filter. The presented experiments, conducted in an indoor scenario with real data, indicate the ability of the proposed method to perform accurate 6 DoF pose estimation.

A State-Space Approach to Dynamic Nonnegative Matrix Factorization

Nonnegative matrix factorization (NMF) has been actively investigated and used in a wide range of problems in the past decade. A significant amount of attention has been given to develop NMF algorithms that are suitable to model time series with strong temporal dependencies. In this paper, we propose a novel state-space approach to perform dynamic NMF (D-NMF). In the proposed probabilistic framework, the NMF coefficients act as the state variables and their dynamics are modeled using a multi-lag nonnegative vector autoregressive (N-VAR) model within the process equation. We use expectation maximization and propose a maximum-likelihood estimation framework to estimate the basis matrix and the N-VAR model parameters. Interestingly, the N-VAR model parameters are obtained by simply applying NMF. Moreover, we derive a maximum a posteriori estimate of the state variables (i.e., the NMF coefficients) that is based on a prediction step and an update step, similarly to the Kalman filter. We illustrate the benefits of the proposed approach using different numerical simulations where D-NMF significantly outperforms its static counterpart. Experimental results for three different applications show that the proposed approach outperforms two state-of-the-art NMF approaches that exploit temporal dependencies, namely a nonnegative hidden Markov model and a frame stacking approach, while it requires less memory and computational power.

Chest-mounted inertial measurement unit for pedestrian motion classification using continuous hidden Markov model

This paper presents a method for pedestrian motion classification based on MEMS inertial measurement unit (IMU) mounted on the chest. The choice of mounting the IMU on the chest provides the potential application of the current study in camera-aided inertial navigation for positioning and personal assistance. In the present work, five categories of the pedestrian motion including standing, walking, running, going upstairs, and going down the stairs are considered in the classification procedure. As the classification method, the continuous hidden Markov model (HMM) is used in which the output density functions are assumed to be Gaussian mixture models (GMMs). The correct recognition rates based on the experimental results are about 95%.

IMU-camera data fusion: Horizontal plane observation with explicit outlier rejection

In this paper, we address the problem of egomotion estimation using an inertial measurement unit and visual observations of planar features on the ground. The main practical difficulty of such a system is correctly determining the ground planar features from the visual observations. Herein, we propose a novel vision-aided inertial navigation system through simultaneous motion estimation and ground plane feature detection. We present a state-space formulation for the pose estimation problem and solve it via an augmented unscented Kalman filter. First, the predictions obtained by the Kalman filter are used to detect the ground plane features. Second, the detected features are fed back to the motion estimation algorithm to be used in the measurement update phase of the filter. The developed detection algorithm consists of two steps, namely homography-based and normal-based outlier rejection. The presented integration algorithm allows 6-DoF motion estimation in a practical scenario where the camera is not restricted to observe only the ground plane. Real-world experiments in an indoor scenario indicate the accuracy and reliability of our proposed method in the presence of outliers and non-ground obstacles.

Observability analysis of a vision-aided inertial navigation system using planar features on the ground

In this paper, we present an observability analysis of a vision-aided inertial navigation system (VINS) in which the camera is downward looking and observes a single point feature on the ground. In our analysis, the full INS parameter vector (including position, velocity, rotation, and inertial sensor biases) as well as the 3D position of the observed point feature are considered as state variables. In particular, we prove that the system has only three unobservable directions corresponding to global translations along the x and y axes, and rotations around the gravity vector. Hence, compared to general VINS, an advantage of using only ground features is that the vertical translation becomes observable. The findings of the theoretical analysis are validated through real-world experiments.