There exists a range of feature detecting and feature matching algorithms; many of which have been included in the Open Computer Vision (OpenCV) library. However, given these different tools, which one should be used? This paper discusses the implementation and comparison of a range of the library's feature detectors and feature matchers. It shows that the Speeded-Up Robust Features (SURF) detector found the greatest number of features in an image, and that the Brute Force (BF) matcher matched the greatest number of detected features in an image pair. Given a benchmark image set, OpenCV's SURF detector found, on average, 1907.20 features in 1538.61 ms, and OpenCV's BF matcher, on average, matched features in 160.24 ms. The combination of the Binary Robust Invariant Scalable Key-points (BRISK) detector and BF matcher was found to be the highest ranked combination of OpenCV's feature detectors and feature matchers; on average, detecting and matching 1132.00 and 80.20 features, respectively, in 265.67 ms. It was concluded that if the number of features detected is important, the SURF detector should be used; else, if the number of features matched is important, the BF matcher should be used; otherwise, the combination of the OpenCV's BRISK feature detector and BF feature matcher should be used.

Color plays an important role in human life: without it life would be dull and monochromatic. Printing color with distinct characteristics, like hue, brightness and saturation, and high resolution, are the main characteristic of image sensing devices. A flexible design of color filter is also desired for angle insensitivity and independence of direction of polarization of incident light. Furthermore, it is important that the designed filter be compatible with the image sensing devices in terms of technology and size. Therefore, color filter requires special care in its design, operation and integration. In this paper, we present a comprehensive review of nanostructured color filter designs described to date and evaluate them in terms of their performance.

Medical exoskeletons are robotic devices that provide paraplegics with the ability to ambulate again. There are four current market leaders in the industry: REX, by Rex Bionics; ReWalk, by Argo Medical Technologies Ltd; Ekso GT, by Ekso Bionics; and Indego, by Parker Hannifin. A weighted table-based method was used to evaluate the capability of these four devices, which was then compared to their maximum speed. The Ekso GT and the ReWalk devices achieved high speeds but had low capability scores due to their instability and minimal safety features. The REX and Indego devices achieved the highest capability scores but for contrasting reasons; the REX devices are inherently stable but have a slow, static gait, while the Indego device relies on walking aids for stability and achieves a faster speed. Both devices scored high in the safety criteria. It was derived from the results that there is an indirectly proportional relationship between speed and stability, and none of the exoskeletons achieve a balance between the two. This study found that there is an opportunity to enhance the speed, mobility, stability, and cost of medical exoskeletons, which is presented below.

Consumer-grade RGB-D cameras capture RGB images along with per-pixel depth information, and because of their limited cost and ability to measure distances at a high frame rate, have been used in robotics and computer vision application. However, drawbacks include the repeatability and accuracy of RGB-D cameras for object detection and localization. This paper investigates and compares RGB-D cameras' performance in terms of depth image quality, depth clouds distribution, etc. performance and configuration methods of frequently used cameras, e.g. PrimeSense, Kinect V1 and Kinect V2, in order to provide useful advice when choosing a camera for robotic applications. Experimental and Point Cloud Library (PCL)-based methods are introduced for point-to-plane distance detection. Based on the obtained results, a relationship between measurements and ground truth is built.

Automated hand gesture recognition is a key enabler of Human-to-Machine Interfaces (HMIs) and smart living. This paper reports the development and testing of a static hand gesture recognition system using capacitive sensing. Our system consists of a $6\times 18$ array of capacitive sensors that captured five gestures—Palm, Fist, Middle, OK, and Index—of five participants to create a dataset of gesture images. The dataset was used to train Decision Tree, Naïve Bayes, Multi-Layer Perceptron (MLP) neural network, and Convolutional Neural Network (CNN) classifiers. Each classifier was trained five times; each time, the classifier was trained using four different participants’ gestures and tested with one different participant’s gestures. The MLP classifier performed the best, achieving an average accuracy of 96.87% and an average $F_1$ score of 92.16%. This demonstrates that the proposed system can accurately recognize hand gestures and that capacitive sensing is a viable method for implementing a non-contact, static hand gesture recognition system.

This paper presents an autonomous method of collecting data for Visible Light Positioning (VLP) and a comprehensive investigation of VLP using a large set of experimental data. Received Signal Strength (RSS) data are efficiently collected using a novel method that utilizes consumer grade Virtual Reality (VR) tracking for accurate ground truth recording. An investigation into the accuracy of the ground truth system showed median and 90th percentile errors of 4.24 and 7.35 mm, respectively. Co-locating a VR tracker with a photodiode-equipped VLP receiver on a mobile robotic platform allows fingerprinting on a scale and accuracy that has not been possible with traditional manual collection methods. RSS data at 7344 locations within a 6.3 × 6.9 m test space fitted with 11 VLP luminaires is collected and has been made available for researchers. The quality and the volume of the data allow for a robust study of Machine Learning (ML)- and channel model-based positioning utilizing visible light. Among the ML-based techniques, ridge regression is found to be the most accurate, outperforming Weighted k Nearest Neighbor, Multilayer Perceptron, and random forest, among others. Model-based positioning is more accurate than ML techniques when a small data set is available for calibration and training. However, if a large data set is available for training, ML-based positioning outperforms its model-based counterparts in terms of localization accuracy.