NEW METHOD FOR IMAGE ANALYSIS USING NEW ALGORITHM FOR CONSTRUCTING VISIBILITY NETWORK IN 3D SPACE

Abstract

The objective of this paper is to determine the input-output relationship between parameters of robot laser cell for hardening, topological property of materials after hardening and volume of material using method of intelligent system; neural network, multiple regression and genetic programming. Used was method for calculate volume of materials after robot laser hardening. The effects and interaction terms on different responses of these selected parameters of robot laser cell for hardening have been analysed using neural network, multiple regression and genetic programming. Comparison of all technique of intelligent system was done. Shown was that the neural network givs the best predicted results. The genetic programming model is better than the regression model. Specimen P16 has the most volume after robot laser hardening, that is 77.7%. Parameter fractal dimension has most impact on regression model and on genetic programming model. It was noticed that topological property of visibility graph effects have considerable influence on the formation of volume, so it cannot be ignored. Used was topological properties of graphs visibility to analyse SEM images of materials after process of robot laser hardening.