Includes bibliography. / The objectives of the project were firstly, to propose and implement the hardware needed for a Machine Vision System that is to operate on the Spiral Ore Concentrator. Secondly, to propose and implement an algorithm that would operate together with the hardware, to find the transition point between the two types of material, also to find the optimum position for the blade separator, while the system operates in a continuous repetitive mode. Finally, to have an on line analysis of the efficiency of the refinement process in the Spiral Ore Concentrator. The hardware of the Machine Vision System was implemented and tested in the laboratory. The hardware included a video recorder, through which a video signal was obtained from the play-back of video material taken of the ore flow in the Spiral Ore Concentrator. The video signal was used as input to an RGB decoder in order to remove the colour modulation. A black and white frame grabber digitized the video signal which was then analyzed by a computer. Based on the fundamental theory of edge detection used in computer vision systems, an algorithm was designed and implemented to detect the transition point between the two types of material. This algorithm was used to find the difference of averages of grey levels between two global neighborhoods within a specified area of interest window. The algorithm gave consistent results and was robust towards surface irregularities. The algorithm, operating in a continuous repetitive mode, gave rapid fluctuations in the determined edge position. Because a motor with a slow response time would be used to control the movement of the blade separator, the determined edge position signal was smoothed by a filter. Based on periodogram analysis of the edge position signal, a smoothing filter was implemented which incorporates a median filter, followed by a fading-memory polynomial filter. These filters gave sufficient smoothing with little lag to step changes in ore concentration. The efficiency of the ore separation was monitored by the determination of losses. These losses consist of, the percentage of black material loss and percentage of white material contamination. Two types of losses could be identified, they were Spraying losses and Filter losses, these were combined to give Total losses. From the true edge position curve, which was obtained by finding the edge position every second pixel point, the Nyquist sampling frequency could be determined. Because of the slow sampling rate, an error in the calculation of the losses was determined from the true edge curve, and the sub-sampled edge curve. Based on this error, it was shown that the machine vision system could multiplex multiple camera inputs.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:uct/oai:localhost:11427/8351 |
| Date | January 1991 |
| Creators | Gold, David |
| Contributors | De Jager, Gerhard |
| Publisher | University of Cape Town, Faculty of Engineering and the Built Environment, Department of Electrical Engineering |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Master Thesis, Masters, MSc |
| Format | application/pdf |
Page generated in 0.0023 seconds