Dương Nguyễn

Trong bài báo này, để giải bài toán chấp nhận tách đa tập (MSSFP) trong không gian Hilbert, chúng tôi trình bày một cách tiếp cận tổng quát để xây dựng các phương pháp lặp. Chúng tôi đề xuất một lược đồ thuật toán xâu trung bình với sự hội tụ yếu và một lược đồ thuật toán xâu trung bình với sự hội tụ mạnh. Lược đồ thuật toán xâu trung bình với sự hội tụ mạnh được xây dựng dựa trên phương pháp lặp tổng quát cho ánh xạ không giãn, trong đó cỡ bước được tính toán trực tiếp trong mỗi bước lặp mà không cần sử dụng chuẩn của toán tử. Những lược đồ thuật toán này không chỉ bao hàm những cải tiến của phương pháp lặp vòng và lặp đồng thời đã biết như những trường hợp riêng mà còn bao hàm cả những phương pháp lặp mới

ABSTRACT This paper presents a simple and robust auto white balance algorithm using the coincidence of chromaticity histograms. After analyzing the relationship between the coincidence of the color histogram and color constancy, the overlap area of chromaticity histograms that keep chromaticity of color images but not the effect of luminance existing in the color histogram is employed to estimate the correct illuminant in scenes. When the overlap reaches the maximum, correspondingly the respective gain coefficients of color channels can be derived to achieve the white balance of the camera. Through numerous experiments and evaluations based on the processing of real world images, the proposed coincidence of chromaticity histograms algorithm achieves the outstanding performance comparing to other algorithms. Furthermore, the simplicity, easy implementation and robustness to the luminance make it flexibly apply to the vision system of the autonomous robot running outdoor.

ABSTRACT Autonomous Ground Vehicle (AGV) has been investigated in large amount researches of robotics but a safe navigation system is still infeasible in complex outdoor environments. One of the biggest challenges is to deal with the presence of vegetation on the vehicle’s way where the decision-making framework usually applied for indoor contexts or rigid objects fails totally. Therefore, this paper addresses a solution for vegetation detection which lets the vehicle fully exploit its mobility capability outside. For that aim, we introduce the use of a new vision system integrated from Photonic Mixer Device (PMD) and CMOS camera, so called Zess-Multicam. Whereby, chlorophyll-rich vegetation is marked by evaluating the reflectance of the modulated near infrared (NIR) given by PMD sensor and the red channel of the CMOS sensor while the color descriptors used also supplement to result a more robust vegetation classifier. Finally we will show the out-performance of this approach in comparison with more conventional approaches under real-time constraint.

ABSTRACT We present a novel approach to explorative path planning for an autonomous mobile outdoor robot. The focus of the proposed method lies on the robust generation of a set of paths that allow the efficient exploration of a previously unknown unstructured outdoor environment. This is achieved by application of a randomized sampling-based path planning approach based on the concept of Rapidly Exploring Random Trees (RRT) used in conjunction with a post-planning analysis of the generated path tree. We introduce the innovative concept of exploration nodes as extension of the basic RRT algorithm. The outcome of the proposed method is a tree of kinematically feasible paths which includes a set of exploration paths that lead to unknown parts of a mobile robot's local environment. This approach is fundamentally different to previously developed RRT-based methods as the principal goal is practical exploration of the unknown instead of planning towards a single local goal. For a complete understanding of the proposed method, we cover the sub-tasks of local map building, path tree generation and path tree analysis within this work. Finally, we demonstrate results of the proposed algorithms from experiments conducted on our autonomous mobile outdoor robot AMOR.