Modal-Based Camera Correction for Large Pitch Stereo Imaging

Motorcycles that See: Multifocal Stereo Vision Sensor for Advanced Safety Systems in Tilting Vehicles

“…Depending on the application, a variety of techniques are used to solve this issue. Examples: (a) In visual odometry it is usual to perform continuous stereo extrinsic re-calibration (5 Degrees of Freedom, DoF) operating on sparse stereo correspondences on stereo frame basis [ 74 ]; (b) In mapping applications the re-calibration is 6 DoF between the cameras in the way of visual odometry but with the addition of GPS information [ 75 ]; (c) In low altitude aerial imagery (< 30 m), the modal deflection of the drone wingspan is monitored which accelerometers in the tip of the wings were the cameras are located, using this information to compensate the relative angle of the stereo pair [ 76 ]; (d) In satellite imagery the undamped micro-vibrations on the satellite are software-compensated by the measures realized over known flat points in the earth [ 77 ]; (e) In areal imagery, a tailored bundle adjustment technique is used to refine camera parameters achieved altitude operations up to 120 m employing a wide baseline [ 78 ]; (f) Automated driving have also bundle adjustment implementations in which they estimate online both extrinsic and intrinsic camera parameters with a pre-definition of the scale [ 79 ]; (g) A recent approach for robotic applications computes 5 DoF of extrinsic by a marker-less nonlinear optimization method [ 80 ]; and (h) In “motion stereo” or Structure-from-Motion (SfM) applications a relaxation of the epipolar constraint is performed. In these cases, the stereo frame is generated for a monocular moving camera, which moves over a rigid scene.…”

Flexible Trinocular: Non-rigid Multi-Camera-IMU Dense Reconstruction for UAV Navigation and Mapping

“…In [11], a down-looking stereo pair with a wide-baseline (0.7 m) is employed and the relative transform between the cameras, together with the poses of the stereo rig itself is estimated in an offline bundle adjustment problem. Lanier et al [12]- [14] use a set of accelerometers distributed along the wing to estimate vibrational disturbances without any vision update. For airborne applications, Yang et al fuse the measurements from a master and slave IMU in an EKF to estimate the relative states, a process referred to as transfer alignment [15].…”

Flexible Stereo: Constrained, Non-Rigid, Wide-Baseline Stereo Vision for Fixed-Wing Aerial Platforms

“…We are aware of two different approaches to real-time baseline estimation in which the baseline deviations closer resemble our scenario: The first approach, described in [14], [15], [16], is based on a modal-analysis of a wing which enables to estimate vibrational disturbances by only measuring with accelerometers (and no gyroscopes). In [14], this method is introduced and show-cased on an Euler-Bernoulli beam model employing two accelerometers. This is extended in [15] to wing-mounted stereo rigs and experimental results are obtained during periodic and random excitation of a real wing model.…”

“…In contrast, we are dealing with an increase of two order of magnitudes (decimeters and more than ten degrees). We are aware of two different approaches to real-time baseline estimation in which the baseline deviations closer resemble our scenario: The first approach, described in [14], [15], [16], is based on a modal-analysis of a wing which enables to estimate vibrational disturbances by only measuring with accelerometers (and no gyroscopes). In [14], this method is introduced and show-cased on an Euler-Bernoulli beam model employing two accelerometers.…”