Multi-threshold Microbial Image Segmentation Based on Improved Firefly Algorithm

Aiming at the problems of large noise interference, uneven gray distribution, and multiple morphologically rich microorganisms in a single image, it is difficult to quickly and accurately extract targets to meet practical application requirements. A multi-threshold segmentation method based on improved firefly algorithm was proposed. Firstly, the optimal multi-threshold number m was automatically obtained by counting the peak value of the gray histogram of the image; Secondly, the two-dimensional entropy threshold segmentation principle was cited,extending two-dimensional entropy single threshold to multiple thresholds anda multi-threshold target cost function was designed based on log entropy; Finally, the intelligent optimization algorithm for the original firefly is prone to fall into the local optimal solution prematurely and the lack of coordination among individual fireflies causes the problem of low algorithm efficiency.An improved firefly algorithm based on the optimization of the firefly initialization process and the adjustment of variable parameters (such as step quantization factor α and relative attractiveness parameter β_ij) was proposed, so as to realize the fast and accurate finding of multiple optimal thresholds. The algorithm was compared with segmentation algorithms such as one-dimensional entropy, twodimensional entropy, Otsu, particle swarm multi-threshold segmentation, and original firefly multi-threshold segmentation, the experimental results show that the accuracy and efficiency of microbial segmentation in this method are significantly improved, laying a good foundation for subsequent automatic identification.