Showing papers on "Orientation (computer vision) published in 2022"

A novel vision-based calibration framework for industrial robotic manipulators

Abstract: With the increasing involvement of industrial robots in manufacturing processes, the demand for high quality robots has increased considerably. A high-quality robot is a robot having good repeatability and accuracy. Industrial robots are known to have very good repeatability, however it is not the same with accuracy. Due to harsh working conditions, accuracy of robots deteriorate over time. Calibration is a practical approach to sustain accuracy. In calibration, position and orientation of the tool center point (TCP) of a robot arm should be corrected using a tracking device with higher accuracy. Different devices such as laser-trackers, optical CMMs, and stereo cameras have been used in the literature. In this paper, a novel calibration framework is proposed based on a single camera and computer vision techniques using ArUco markers. The product of exponentials method is used for kinematic modeling of the robot to avoid the singularity. The performance of the framework is tested using computer-based simulations and using a six degree of freedom (6-DOF) UR5 robotic manipulator. Position and orientation errors are used as metrics in the experiments. The position and orientation errors in real world experiments reached to 2.5 mm and 0.2°, respectively. The result shows that the method is usable in real world scenarios.

Robust Image Encryption and Zero-watermarking Scheme Using SCA and Modified Logistic Map

Abstract: In this work, we first present a modified version of the traditional logistic chaotic map. The proposed version contains an additional parameter that is used to increase the security level of the proposed digital image copyright protection scheme. The latter merges two methods of image copyright protection, namely the image zero-watermarking and image encryption, which provides a high level of security when communicating images via the Internet. Next, we discuss the influence of geometric attacks on the efficiency of the proposed scheme, and then we introduce an efficient solution that can resist such attacks. The proposed solution involves the use of Sine Cosine Algorithm (SCA) with an appropriate algorithm suitable for the correction of geometric attacks (image translation, orientation and its combination) applied to the encrypted image. On the one hand, the simulation results show that the proposed scheme provides a high level of security and can resist various attacks (differential, common image processing, geometric, etc.). On the other hand, the conducted comparison in terms of robustness against geometric attacks clearly demonstrates the superiority of our scheme over recent image encryption ones.

Learning conditional photometric stereo with high-resolution features

Abstract: Photometric stereo aims to reconstruct 3D geometry by recovering the dense surface orientation of a 3D object from multiple images under differing illumination. Traditional methods normally adopt simplified reflectance models to make the surface orientation computable. However, the real reflectances of surfaces greatly limit applicability of such methods to real-world objects. While deep neural networks have been employed to handle non-Lambertian surfaces, these methods are subject to blurring and errors, especially in high-frequency regions (such as crinkles and edges), caused by spectral bias: neural networks favor low-frequency representations so exhibit a bias towards smooth functions. In this paper, therefore, we propose a self-learning conditional network with multi-scale features for photometric stereo, avoiding blurred reconstruction in such regions. Our explorations include: (i) a multi-scale feature fusion architecture, which keeps high-resolution representations and deep feature extraction, simultaneously, and (ii) an improved gradient-motivated conditionally parameterized convolution (GM-CondConv) in our photometric stereo network, with different combinations of convolution kernels for varying surfaces. Extensive experiments on public benchmark datasets show that our calibrated photometric stereo method outperforms the state-of-the-art.

Histograms of oriented mosaic gradients for snapshot spectral image description

Abstract: This paper presents a feature descriptor using Histogram of Oriented Mosaic Gradient (HOMG) that extracts spatial-spectral features directly from mosaic spectral images. Spectral imaging utilizes unique spectral signatures to distinguish objects of interest in the scene more discriminatively. Snapshot spectral cameras equipped with spectral filter arrays (SFAs) capture spectral videos in real time, making it possible to detect/track fast moving targets based on spectral imaging. How to effectively extract the spatial-spectral feature directly from the mosaic spectral images acquired by snapshot spectral cameras is a core issue for detection/tracking. So far, there is a lack of comprehensive and in-depth research on this issue. To this end, this paper proposed a new spatial-spectral feature extractor for mosaic spectral images. The proposed scheme finds two forms of SFA neighborhood (SFAN) to construct a feature extractor suitable for any SFA structure. Exploiting the spatial-spectral correlation in two SFANs, we design six mosaic spatial-spectral gradient operators to compute spatial-spectral gradient maps (SGMs). HOMG descriptors are constructed using the magnitude and orientation of SGMs. The effectiveness and generalizability of the proposed method have been verified with object tracking experiments. Compared to the state-of-the-art feature descriptors, HOMG ranked first on two datasets captured with snapshot spectral camera with different SFAs, achieving a gain of 3.9% and 5.9% in average success rate over the second-ranked feature.

Deep neural models for automated multi-task diagnostic scan management-quality enhancement, view classification and report generation.

Abstract: The detailed physiological perspectives captured by medical imaging provides actionable insights to doctors to manage comprehensive care of patients. However, the quality of such diagnostic image modalities is often affected by mismanagement of the image capturing process by poorly trained technicians and older/poorly maintained imaging equipment. Further, a patient is often subjected to scanning at different orientations to capture the frontal, lateral and sagittal views of the affected areas. Due to the large volume of diagnostic scans performed at a modern hospital, adequate documentation of such additional perspectives is mostly overlooked, which is also an essential key element of quality diagnostic systems and predictive analytics systems. Another crucial challenge affecting effective medical image data management is that the diagnostic scans are essentially stored as unstructured data, lacking a well-defined processing methodology for enabling intelligent image data management for supporting applications like similar patient retrieval , automated disease prediction etc. One solution is to incorporate automated diagnostic image descriptions of the observation/findings by leveraging computer vision and natural language processing. In this work, we present multi-task neural models capable of addressing these critical challenges. We propose ESRGAN, an image enhancement technique for improving the quality and visualization of medical chest x-ray images, thereby substantially improving the potential for accurate diagnosis, automatic detection and region-of-interest segmentation. We also propose a CNN-based model called ViewNet for predicting the view orientation of the x-ray image and generating a medical report using Xception net, thus facilitating a robust medical image management system for intelligent diagnosis applications. Experimental results are demonstrated using standard metrics like BRISQUE, PIQE and BLEU scores, indicating that the proposed models achieved excellent performance. Further, the proposed deep learning approaches enable diagnosis in a lesser time and their hybrid architecture shows significant potential for supporting many intelligent diagnosis applications.

3D reconstruction and positioning of surface features based on a monocular camera and geometric constraints

Abstract: This paper proposes a 3D reconstruction and positioning method based on a single detecting camera and geometric constraint to evaluate the defect quantitatively. We realize the camera orientation through datum point coordinates acquired by a photogrammetry system and then solve the equation of the defect detection camera model and surface constraints to reconstruct the color point cloud in 3D space. To realize the reconstruction and orientation of defects on objects without marker points, the system is operated by a repetitive moving of robotic arms. Finally, the experiment achieves the reconstruction of three types of products with marker points and one of the same types of objects without marks. The experiment shows that this method can limit the accuracy of surface feature positioning to 0.7 mm, and relative accuracy of the area calculation to 0.125%.

Stereophotogrammetry Camera Pose Optimization

Abstract: Stereophotogrammetry makes use of calibrated camera pairs to obtain three-dimensional information from two-dimensional images. The accuracy of the extracted measurements is extremely dependent on the selection and setup of the camera system. For a given test object and desired viewing orientation, there is no one “correct” stereo camera setup, but rather a range of potential setups with some approaching an optimal system with respect to maximizing the measurement resolution. The open-ended nature of this test design exercise is compounded by equipment availability and the fact that many of the setup parameters have dependent characteristics, e.g., changing focal distance will affect stand-off distance, field of view, and image projection, among others. This work describes a planning tool that utilizes projective and Euclidian geometry to iteratively estimate optimal camera poses for available equipment, determines the most efficient image size, and also performs checks for lens diffraction, minimum focal distance, and adequate depth of field. Integrating a finite element model with these calculations further extends planning capabilities by allowing (1) an accurate definition of the volume to be imaged and (2) the ability to estimate response displacements in pixels due to an arbitrary excitation applied to the test object. This latter capability is critical for pre-test determination of the chosen camera setup’s ability to successfully extract three-dimensional measurements. The theory and workflow are presented along with an experimental demonstration.