Improved whale optimization algorithm for 2D-Otsu image segmentation with application in steel plate surface defects segmentation

Since the steel plate surface defect image often has complicated background and lots of noise, the segmentation accuracy is low when using the single threshold Otsu method. Therefore, this paper introduces the whale optimization algorithm (WOA) to optimize the threshold of the dual-threshold image segmentation. To avoid the premature convergence, slow convergence speed and easy fall into the local optimum of the original WOA, an improved WOA is proposed. Firstly, the WOA is discretized by using round function; secondly, the sin mapping generation chaotic sequence is used to replace the randomly generated initial population in the initialization process of the WOA to enhance the multiformity of population; thirdly, the global search and local development capabilities are balanced and improved by nonlinear time-varying factors and inertia weights in the position updating mechanism; finally, the improved WOA is applied to the two-dimensional Otsu (2D-Otsu) algorithm to select the optimal threshold for image segmentation. The simulation results of 8 classic benchmark functions show that the improved WOA can obtain the optimal value of the function 0, - 12,569.5. The improved WOA can raise convergence speed and improve the global search ability and get rid of the local optimum. The experimental results show that the proposed algorithm outperforms the Otsu algorithm and can achieve more accurate segmentation of steel plate surface defect image. Compared with 2D-Otsu algorithm, the proposed algorithm reduces running time by 0.34 s and has the highest segmentation efficiency for rolled-in scale defects.