S. Y. Chen

Kalman Filter for Robot Vision: A Survey

Kalman filters have received much attention with the increasing demands for robotic automation. This paper briefly surveys the recent developments for robot vision. Among many factors that affect the performance of a robotic system, Kalman filters have made great contributions to vision perception. Kalman filters solve uncertainties in robot localization, navigation, following, tracking, motion control, estimation and prediction, visual servoing and manipulation, and structure reconstruction from a sequence of images. In the 50th anniversary, we have noticed that more than 20 kinds of Kalman filters have been developed so far. These include extended Kalman filters and unscented Kalman filters. In the last 30 years, about 800 publications have reported the capability of these filters in solving robot vision problems. Such problems encompass a rather wide application area, such as object modeling, robot control, target tracking, surveillance, search, recognition, and assembly, as well as robotic manipulation, localization, mapping, navigation, and exploration. These reports are summarized in this review to enable easy referral to suitable methods for practical solutions. Representative contributions and future research trends are also addressed in an abstract level.

The Specular Exponent as a Criterion for Appearance Quality Assessment of Pearllike Objects by Artificial Vision

For pearls and other smooth alike lustrous jewels, the apparent shininess is one of the most important factors of beauty. This paper proposes an approach to automatic assessment of spherical surface quality in measure of shininess and smoothness using artificial vision. It traces a light ray emitted by a point source and images the resulting highlight patterns reflected from the surface. Once the reflected ray is observed as a white-clipping level in the camera image, the direction of the incident ray is determined and the specularity is estimated. As the specular exponent is the most important reason of surface shininess, the method proposed can efficiently determine the equivalent index of appearance for quality assessment. The observed highlight spot and specular exponent measurement described in this paper provide a way to measure the shininess and to relate the surface appearance with white-clipped image highlights. This is very useful to industrial applications for automatic classification of spherical objects. Both numerical simulations and practical experiments are carried out. Results of objective and subjective comparison show its satisfactory consistency with expert visual inspection. It also demonstrates the feasibility in practical industrial systems.

Robust 3D Shape Reconstruction from a Single Image Based on Color Structured Light

Reconstructing 3D shapes from 2D images based on structured light is becoming an increasingly important topic in computer vision. However, low resolution and sensitive to environment illumination are the main restriction of this technology for practical application. This paper proposes a new color coded structured light technique for reconstructing object shape from a single image. This technique works by projecting a pattern of color stripes with white gaps and assigning the projected stripe color by color classification. The method can enhance the robustness with respect to uncontrolled environment illumination and color cross-talk. The stripe boundary is located accurately by local searching method. Additionally, we propose a technique to achieve dense shape reconstruction by shifting the same patterns. Practical experimental results are provided to demonstrate the performance of the proposed methods. Furthermore, 3D models with high quality and resolution were produced under uncontrolled light condition.

A Self-Calibration Method for Rotational Stereo Vision

Self-calibration for imaging sensors is essential to many computer vision applications. In this paper, we proposed a simple self-calibration method of a camera with constant internal parameters for rotational stereo vision. What we mainly utilize is the properties of the scene geometry. The basis of our approach is to calculate the intersections between the vanishing line and the track of the corresponding 3-D points in different images to obtain the images of the circular points. Subsequently, the image of absolute conic can be identified and then the calibration matrix of the camera can be calculated. Experiments on both synthetic and real image sequence are given to show the accuracy of this algorithm.

Point Cloud Simplification Based on an Affinity Propagation Clustering Algorithm

Point cloud simplification is an important step in reverse engineering and computer vision. Nowadays many researchers are directly working on point sets other than polygonal meshes, while some nasty problems still exist, such as time cost, memory cost and accuracy. This paper proposes a novel method for point cloud simplification by integrating both re-sampling and Affinity Propagation Clustering. The advantage of Affinity Propagation clustering is passing messages among data points and fast speed of processing. Together with the iterative re-sampling, it can dramatically reduce the duration of the process while keep a lower memory cost. The results of simulative experiments demonstrate that the proposed algorithm outperformed traditional clustering or re-sampling methods.

Entire Object Reconstruction Using a Rotational Stage

This paper presents a method for reconstruction of entire three-dimensional objects based on principles of the rotational stereo. The experimental system, which has three degrees of freedom (two for translation and one for rotation), uses an accurate stage to make all areas of the object surface visible. It only needs one camera to capture images and thus reduces the implementation cost. The object to be reconstructed is placed on a rotational device which can be precisely controlled by a computer. A series of images are easy to be obtained for recovering the complete model of the object. A few examples are carried out in the experiments. Results are very satisfactory in mean of both feasibility and accuracy.