Direct Depth Estimation Research Articles

Fuzzy Decision Tree Based Characterization of Subsurface Anomalies

Recent intervention of machine learning based methodologies into infrared thermography proves to provide better defect detection and characterization. This paper provides a Fuzzy decision tree based quantitative post processing modality along with a theoretical model for thermal waves to characterize the subsurface anomalies using quadratic frequency modulated thermal wave imaging. A carbon fiber reinforced polymers (CFRP) and mild steel (MS) specimens having flat bottom holes with different depths and sizes are processed through experimental evaluation. A direct depth estimation is provided by the proposed modality being evaluated from the proposed mathematical model. In addition, its detection capability and reliability over other contemporary approaches is compared using signal to noise ratio, defect sizing and probability of defect detection.

Deep monocular 3D reconstruction for assisted navigation in bronchoscopy.

In bronchoschopy, computer vision systems for navigation assistance are an attractive low-cost solution to guide the endoscopist to target peripheral lesions for biopsy and histological analysis. We propose a decoupled deep learning architecture that projects input frames onto the domain of CT renderings, thus allowing offline training from patient-specific CT data. A fully convolutional network architecture is implemented on GPU and tested on a phantom dataset involving 32 video sequences and [Formula: see text]60k frames with aligned ground truth and renderings, which is made available as the first public dataset for bronchoscopy navigation. An average estimated depth accuracy of 1.5mm was obtained, outperforming conventional direct depth estimation from input frames by 60%, and with a computational time of [Formula: see text]30ms on modern GPUs. Qualitatively, the estimated depth and renderings closely resemble the ground truth. The proposed method shows a novel architecture to perform real-time monocular depth estimation without losing patient specificity in bronchoscopy. Future work will include integration within SLAM systems and collection of in vivo datasets.

Depth estimating Full Tensor Gravity data with the Adaptive Tilt Angle method

Depth estimation procedures for potential field data are well recognised techniques. Both Euler and Werner methodologies are typically used as a series of automated steps and applied to both gridded and profile data. The Tilt Derivative Depth method works on gridded data and has been used extensively on magnetic data. Its advantage is its ability to produce a focused set of solutions and is now being commonly adopted for potential field data. This paper describes an Adaptive Tilt Angle method for depth estimating Full Tensor Gravity data. The method is an adaptation of the Tilt Derivative depth estimation procedure adopted for magnetic data. The procedure works on 4 of the independently measured Tensor components and produces sets of solutions that are more easily interpreted. The tilt angle method is defined as a ratio of the Tensor components in each of the X, Y and Z directions and assumes a vertical contact geological setting. The implementation of a scaling factor allows the technique to work on horizontal contacts. The scaling factor is essentially similar to the concept of a Structural Index as used with Euler depth estimation methods. The technique was tested successfully on an Air-FTG® survey data set over a shallow salt feature onshore USA and is now being routinely deployed. The benefits of the direct depth estimation technique are immense in that it not only provides constraint on other interpretative processing techniques, but quickly establishes a starting depth model for any detailed forward / inverse modelling exercises.

Adaptive control of nonlinear visual servoing systems for 3D cartesian tracking

This paper presents a control strategy for robot manipulators to perform 3D cartesian tracking using visual servoing. Considering a fixed camera, the 3D cartesian motion is decomposed in a 2D motion on a plane orthogonal to the optical axis and a 1D motion parallel to this axis. An image-based visual servoing approach is used to deal with the nonlinear control problem generated by the depth variation without requiring direct depth estimation. Due to the lack of camera calibration, an adaptive control method is used to ensure both depth and planar tracking in the image frame. The depth feedback loop is closed by measuring the image area of a target object attached to the robot end-effector. Simulation and experimental results obtained with a real robot manipulator illustrate the viability of the proposed scheme.

Simple time-domain optical method for estimating the depth and concentration of a fluorescent inclusion in a turbid medium

A simple time-domain optical method for estimating the depth and concentration of fluorescent inclusions in turbid media is described. We demonstrate direct depth estimation of a localized fluorescent object from the temporal position of the temporal point-spread function maximum. The depth estimation permits recovery of the fluorophore concentration, both of which are essential quantities for optical molecular imaging studies. Since the maximum is independent of the fluorophore concentration, excitation laser power, detector gain, and other system-dependent factors, this method ensures a robust and efficient approach.

A Kalman filter approach to direct depth estimation incorporating surface structure

The problem of depth-from-motion using a monocular image sequence is considered. A pixel-based model is developed for direct depth estimation within a Kalman filtering framework. A method is proposed for incorporating local surface structure into the Kalman filter. Experimental results are provided to illustrate the effect of structural information on depth estimation.

Depth estimation from a sequence of monocular images with known camera motion

This paper reports an approach of computing depth maps from a monocular image sequence, under the assumption that the camera motion is known. The direct depth estimation method is combined with the optical flow based method to improve estimation accuracy. The optical flow on and near moving edges are computed using a correlation technique. The optical flow information is then fused with gradient information to estimate depth not only on the moving edges but also in the internal regions. The depth estimation problem is formulated as a Kalman filter problem and is solved in three stages. In the prediction stage, the depth map estimated for the current frame, together with knowledge of the camera motion, is used to predict the depth and depth variance at each pixel in the next frame. In the estimation stage, a Kalman filter is employed to refine the predicted depth map. The resulting estimation algorithm takes into account the information from the neighboring pixels. In the smoothing stage, based on the error covariance information, morphological filtering is applied to reduce the effect of measurement noise and to fill in untrustable areas. Simulation results are provided to demonstrate the effectiveness of the proposed method.