Brendan Khoo

A design methodology of a flexible robotic arm vision system for OTOROB

Telemedicine is a new and promising application in the field of medical science. Medical consulting and remote medical procedures or examinations became more effective with the development of interactive audiovisual systems and remote mobile robotic platforms. Tele-presence requires real time video transmission. Fixed mounting of video device on mobile robotic platform limits the video projection at certain view angle, thus restricting the remote doctor in obtaining the fine and accurate visual information of the patient. To circumvent the problem, this paper presents a flexible robotic arm with vision system for OTOROB (Orthopedic Robot). A four degree of freedom (DOF) robotic arm with extending and retracting feature enables the remote doctor to articulate the attached video device on robotic arm to target area on patient. The working envelope of the robotic arm increased with the linear motion feature. This combined with yaw and pitch at the end effectors improves the visual projection at the intended target point.

Probabilistic multi SVM weed species classification for weed scouting and selective spot weeding

In this paper, a probabilistic output multi SVMs were used to classify the weed seedlings into groups for spot spraying and weed scouting application. Weeds in the samples are collected at approximely 1-4 weeks after post emergence. The weed seedlings are classified using Support Vector machines while feature selection and fine tuning of classifier parameters were fine tuned using genetic algorithm. The features which included regional shapes parameters, fractal dimensions and elliptical Fourier coefficients, skeleton statistics, boundary to centroid and colour statistics were extracted from individual leaves and the overall binarized shape of the weed seedlings. The resulting SVM ensemble classifier is able to classify the various weed seedlings into various classes at a reasonable rate which can be further improved by enlarging training sets and improving individual SVMs.

Simultaneous rotation and translation movement for four omnidirectional wheels holonomic mobile robot

Navigation and positioning is the main system for a mobile robot. Holonomicity in the field of robotic has becoming more and more important as it allow the mobile robot to do translation and rotation movements. Although many research had done to overcome the poor motion stability and trajectory motion using three omnidirectional wheels, they may have stability problems due to triangular contact area with the ground and the payload they carry. Thus, to increase the movement efficiency of the holonomic mobile robot, simultaneous translation and rotation movements for four omnidirectional wheels holonomic mobile robot was introduced. To create the system for the movement, a methodology is laid out in this paper. A kinematic analysis was done on the movement of the holonomic mobile robot. The algorithm of the kinematic movement of the mobile robot then was developed. A test grid was carry out to test the navigation algorithm of the system and the result is within the acceptable range. This algorithm would allow the holonomic mobile robot to do translation and rotation movements either simultaneously or independently.

Detection of Amaranthus palmeri sp. Seedlings in Vegetable Farms Using Genetic Algorithm Optimized Support Vector Machine

This research is a part of a larger research scope to recognise individual weed species for weed scouting and spot weeding. Support Vector Machines are used to classify the presence of specified weeds (Amaranthus palmeri ) by analysing the shape of the weeds. Weed leaves are extracted using image dilation and erosion methods. Several shape feature types were proposed and a total of 59 features were used as the feature pool. The feature selection and fine tuning of the Support Vector Machine are performed using Genetic Algorithm. The outcome is a generalised classifier that enables classification of weed leaves with an average of 90.5% classification rate.

An Application on Medical Tele-Diagnosis Robot (MTR) for Real-Time Motion Detection

The use of medical robots in healthcare industry especially in rural areas are hitting limelight these days. Development of Medical Tele-diagnosis Robot (MTR) has gain importance to unravel the need of medical emergencies. Nevertheless, challenges for a better visual communication still arises. Thus, a face identification and tracking system for MTR is designed to allow an automated visual which will ease the medical specialist to identify and keep the patient in the best view for visual communication. This paper emphasis on the motion detection module which is the first module of the system. An improved motion detection technique is proposed which suits a real-time application for a dynamic background. Frame differencing method was used to detect the motion of the target. The developed motion detection module succeeded an accuracy of 96% resulting an average of 97% of the whole MTR.