This paper presents the model analysis of ball-on-sphere system by considering the effect of friction. The ball-on-sphere system is modelled using bond graph technique. In the bond graph modelling procedures of the system, the various subsystems, storage elements, junction structures, transformer elements, dissipating element with appropriate causality assignments and energy exchange that make up the ball-on-sphere system were identified and modelled. In the model analysis of the ball-on-sphere system, the developed model with effect of friction had time of angular position response of the ball ( ) achieved at 0.5253s while in the system model without effect of friction, time of 0.5408s was achieved for the angular position response of the ball ( ). This shows 2.9% improvement of the angular position response of the ball considering frictional effect in the developed bond graph model of the system.

In this research work, an improved active contour method called Bat-Active Contour Method (BA-ACM) using bat algorithm has been developed. The bat algorithm is incorporated in order to escape local minima entrapped into by the classical active contour method, stabilize contour (snake) movement and accurately, reach boundary concavity. Then, the developed Bat-Active Contour Method was applied to a dataset of medical images of the human heart, bone of knee and vertebra which were obtained from Auckland MRI Research Group (Cardiac Atlas Website), University of Auckland. Set of similarity metrics, including Jaccard index and Dice similarity measures were adopted to evaluate the performance of the developed algorithm. Jaccard index values of 0.9310, 0.9234 and 0.8947 and Dice similarity values of 0.8341, 0.8616 and 0.9138 were obtained from the human heart, vertebra and bone of knee images respectively. The results obtained show high similarity measures between BA-ACM algorithm and expert segmented images. Moreso, traditional ACM produced Jaccard index values 0.5873, 0.5601, 0.6009 and Dice similarity values of 0.5974, 0.6079, 0.6102 in the human heart, vertebra and bone of knee images respectively. The results obtained for traditional ACM show low similarity measures between it and expertly segmented images. It is evident from the results obtained that the developed algorithm performed better compared to the traditional ACM.