Yeonho Lee

Wide-baseline stereo matching based on the line intersection context for real-time workspace modeling.

Line matching in widely separated views is challenging because of large perspective distortion and violation of the planarity assumption in local regions. We introduce a novel method of wide-baseline image matching based on the coplanar line intersections for poorly textured and/or nonplanar structured scenes. The local areas of the coplanar line pairs are normalized into canonical frames by rectifying the coplanar line pairs to be orthogonal. Then, the 3D interpretation of the intersection context of the coplanar line pairs helps to match the nonplanar local regions. Furthermore, for calibrated stereo cameras, we propose a matching criterion based on 3D planar homography to improve the matching accuracy while reconstructing most likely physically existing planar patches. Experimental results demonstrate the effectiveness of the proposed method for real-world scenes.

Structured light camera base 3D visual perception and tracking application system with robot grasping task

3D vision-based recognition as well as grasping of complex objects is required not only for detection and categorization but also for pose estimation and robotic pick-and-place operations. In this paper, we propose a structured light camera based 3D visual perception and tracking system application with robot arm grasping for manufacturing. In the first step, we use a geometric surface primitive patch segmentation approach based on Hough transforms to obtain accurate surface normal estimations from 3D point clouds for the identification of patch primitives. The most relevant primitives for our application include planar and cylindrical surface patches. We extract primitive surface patches from automotive CAD models in DXF or 3DS format. The models are then decomposed to simple entities such as planar polygons, vertexes and lines. Our system takes advantage of the available CAD data for both object recognition and for pose estimation. In the final step, we propose point based KLT tracking method with dominant features from object recognition results both raw data and CAD. Our experimental results demonstrate that we can show application results from YASKAWA MOTOMAN robots demonstration. Whole processes perform within few seconds, an accurate pose and reasonable tracking.

Accurate and fast extraction of planar surface patches from 3D point cloud

Planar surface patches, extracted for instance from a captured 3D point cloud, play an important role as a feature in modeling and/or recognizing objects and environments. In terms of extracting such planar surface patches, the main technical issue involved may be of overcoming the trade-off between the modeling accuracy and the computational efficiency, especially, when the 3D point cloud captured is noisy. Conventionally, they have taken approaches that seek either for high computational efficiency at the expense of modeling accuracy or vice versa. This paper contributes the advancement in the methodology of extracting planar surface patches from a noisy 3D point cloud by presenting a method that provides high modeling accuracy yet under low computational cost. The major contribution of the proposed method is summarized as follows: 1) A robust estimation of surface normal vectors in such a way as to minimize the effect of noise in the data. 2) An accurate localization of density peaks based on a spherical coordinate operator with a flexible size of sliding window used for estimating the density of surface normal vectors represented on a unit spherical surface as a means of accurately identifying those planar surface patches of the same orientation. 3) A segmentation of individual planar surface patches of the same orientation by projecting those 3D points clustered as belonging to a same peak onto x, y and z axes of the Cartesian coordinate with the Y axis representing their orientation. The experimental results show the effectiveness of the proposed method in comparison with conventional ones for various indoor and outdoor images.

Improved industrial part pose determination based on 3D closed-loop boundaries

Although various pose estimation methods for industrial robots have been studied, there are few methods that can be applied to real industrial environments. The reason is that previous methods cannot be used for a wide range of industrial parts, or the methods have many limitations in a real industrial factory. In this paper, we break through the weakness of the previous 3D closed-loop boundary based industrial part pose estimation method, which can be extracted in the range image provided by a 3D scanner. After extracting the 3D closed-loop boundaries, we make a group with 3D closed-loop boundary sets. Each 3D closed-loop boundary generates the three principal components (eigenvalues and eigenvectors), the centroid of the boundary and topological relationship in the set. Since this proposed method can be applied to the perfect circle 3D closed-loop boundary in industrial parts, it can be employed for more various general-purpose applications. Through experiments on industrial objects and industrial robots, we validate that the proposed method is feasible for real factory environments.

Optical Encryption of a Binary Image by Phase Modulation of the Wavefront

We present a new scheme for optical encryption of a binary image. In our method, the original binary data page is first divided into two identical pages. In each data page, the “on” and “off” pixels are represented by two discrete phases that are 90° apart. The first page corresponds to the phase conjugation of the second page, and vice versa. In addition, the wavefront of the two data pages is changed simultaneously from planar to spherical, for better encryption. The wavefront modification is represented by an extra phase shift, which is a function of position on the wavefront. In this way the two separate pages are both encrypted, and therefore the pages cannot be distinguished in a CCD. If the first page is used as an encrypted data page, then the second page is used as the decryption key, and vice versa. The decryption can be done by simply combining the two encrypted data pages. It is shown in our experiment that encryption and decryption can be fully accomplished in the optical domain.

Characterization of IPS-LCD by Reduced One-dimensional Model

Based on the distinctive characteristics observed in the intensity transmittance of an IPS-LC panel, the previous one-dimensional model is greatly reduced such that only a few data points and their interpolations predict the intensity transmittance of an IPS-LCD with a small error for arbitrary gray levels. Experimental procedure and numerical methods are described in detail.

Structural Characteristics of a Conical-Frustum-Patterned Stretchable Heater in an External-Force Environment

Stretchable heaters are mechanically and biomedically advantageous because they are more flexible than the conventional microheaters. To determine the way that a heater can be used appropriately in a wide range of application fields, studies on the mechanical and thermal characteristics regarding various film configurations are underway. In this study, the mechanical characteristics of conical-frustum-patterned stretchable heaters with various configurations with upper-radius (R1), bottom-radius (R2), and height (h) aspect ratios were numerically investigated when tensile forces and bending moments were applied to the film of heater. The temperature and von Mises stress of the film were analyzed using the commercial software COMSOL ver. 5.2. The structural characteristics of a conical-frustum-patterned stretchable heater were evaluated under external load, and results are graphically depicted.

Object recognition and pose estimation based on 3D planar closed loop boundaries

Industrial parts often have various closed loop boundaries (CLBs) of a convex geometric shape such as bolt holes. CLBs can serve as an effective 3D geometric feature for the recognition and pose estimation of industrial parts. However, an accurate extraction of 3D CLBs from a single shot 3D point cloud has been hampered by the possible CLB self-occlusion from a 3D camera perspective, resulting in CLB data separation. In this paper, we present an approach to solving such a self-occlusion issue for accurately identifying 3D CLBs on a planar surface patch. The proposed approach removes out the effect of self-occlusion by transforming an original single shot 3D point cloud represented with reference to the camera frame into that with reference to a frame z-axis of which is orthogonal to the CLB surface patch. This allows the 3D edge of a CLB to be extracted without data separation. Experimental results show that the proposed method not only enable 3D CLB extraction for the case of an arbitrary camera tilting but also improves the accuracy over a conventional method by 30% even for the case of no camera tilting.

Mechanical characteristics of stretchable electronics based on a mogul-patterned structure

Recent progress in unconventional forms of foldable and stretchable electronics has enabled development of novel electronics. Especially, stretchable electronics have attracted much interest for applications that require reliable operation under mechanical deformation. Stretchable electrodes can be bent, stretched and compressed while maintaining good performance and structural safety. To realize flexible electronics, stretchability of devices, which requires a deeper understanding of nanoscale materials and mechanics, is needed. In this study, the mechanical characteristics of electrodes based on a mogul-pattern, which retain their electrical conductivity under mechanical deformation, were evaluated. Nanostructured patterns and thin film material were considered for stretched and bent models. The von Mises stress and strain distribution were analyzed and depicted graphically.

Holography optical memory recorded with error correcting bits

A novel error correction method is proposed for volume holographic memory systems. In this method the information of two adjacent binary bits is recorded in the space between the two bits, which is used to correct the errors in the data bits. The new method is compared with (15, 5) Reed—Solomon code using the same redundancy of 200% as ours. It is shown that the new method achieves the similar bit error rate as the RS code.