Detektiranje in razpoznavanje predmetov na vodnih površinah s pomočjo izločanja barvnih območij : doktorska disertacija

slovenian

doktorska dela

2013

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

This thesis covers the issues of detection and recognition of objects on watersć surfaces based on color elimination. The goal of doctoral thesis is todevise a system that enables us detection and recognition of objects on watersć surfaces in the vicinity of energy facilities, hydro power plants. Such a system allows us to detect risks of objects, that represent a life danger, vessels, sport activities in this area such as boating, rafting, swimming, driving with motorboats and others on time. System for detection andrecognition of objects on watersć surfaces is also capable of detecting drift-wood, which like vessels, can present a serious problem in the vicinity of hydro power plants. A system that enables us detection and recognition of objects on watersć surfaces can also be used on other fields, if used by the system introduced in thesis. The system for detection and recognition objects on waterćs surfaces first of all needs appropriate selective image material onwhich the whole detecting system will work. Idea of detecting objects on watersć surfaces presented in thesis, uses a shade of blue and brown color forits base, which has been, according to given image base the best choice for further detection. For successful working of the system in the very beginning we have to carry out image pre-processing and define suitable size of individual image and choose proper ones for further use from a number. Further the system defines color area, which is specific for a fixed field of system working. Because all further steps of detection then refer to correct definition of color area with this system we can in the very beginning define field or options of detection system working which uses this system. In case of using this system in any of the other fields we have to use different colorparameters and define basic color area, which we will successfully eliminate in further procedure. Moreover the system defines color deviation with help of Euclidean distance. Regarding to entering demands, parameters andedge conditions which are required by system itself a defined image is processed according to individual picture element - ĆpixelĆ. In this way the system examines color composition of particular pixel and memorizes color values of three basic colors: red, green and blue. That is how we get RGB (ĆRed Green BlueĆ) values for each individual element and use them in definingcolor deviation. After the procedure of defining color deviation with Euclidean distance we get a chart of values of individual pixel. That is followed by transformation of an image in grey color area. Then the system anticipates a definition of thresholding border. Similarly as in defining color area we also experimentally define thresholding border according to the field of system usage and it depends on level of necessity of eliminating defined color area of specific image. All values of individual pixels that do not satisfy border criteria and do not fall into specific color area are eliminated from further processing and are considered as useless. Next is creating a morphological variable with which we can later on in the procedure of eroding specific color segments easily detect elements in the image that actually present objects for detection. Form, length and direction of the morphological variable itself are experimentally defined to at the most satisfy edge conditions which are set by detection system. After the eroding procedure comes a review of pixels that satisfied edge conditions above the base image. Above the original image we get a created mask that contains elements of color area defined in the beginning. With elimination of this areawe extract an object we wanted to detect from original image.

07.08.2023