Saturated Iteration of Neighborhood Averaging Filter Algorithm For Kinect Structural Noise Eliminator For Wild Thumper

Abstract

A contemporary study (Andersen, 2012) on the Kinect sensor as a visual sensor device for robots explains that the sensor has some fundamental flaws. Among them there is a shadow on the results of sensing edge objects that will affect the process of recognition of shape (shape recognition) spatially. If the Kinect sensor is used in the robot vision navigation system, the sensor may lead to errors in the robot's decision on the shape of the object being sensed. The experiment was aimed to measure the saturation iteration level for the Localized Neighborhood-Averaging method in its effect on the shadow thickness (structural noise) produced by kinect sensors. Robot model that will be used for testing is the robot Wild Thumper 4WD. The kinect sensor is conditioned to capture a square black object measuring 10cm x 10cm perpendicular to a homogeneous background (white with RGB code 255,255,255). The result of kinect sensor data will be taken using Visual Basic 6.0 program periodically by repeating the Localized Neighborhood-Averaging method 1x up to 5x iteration. Observation will be done by looking at the similarity of edge slices of objects captured by Kinect sensors. Saturated iteration Neighborhood-Averaging filter on pixel noise generated by kinect is on the 3rd iteration based on observation of edge section with 2 iteration ratio showing no change in noise thickness. The noise thread at the 2nd and 3rd iterations has the same contours based on the Coefficient of Variation calculations applied to both data. The use of filters is by filter masking 3x3 matrix size. It is expected that further research can be developed on the size of 5x5 or more masking filters in the hope of having a better saturation iteration tolerance result.