Detection and distinguishing between different species of bacteria using experimental microbiology is an expensive, time-consuming, and risky process. Automatic computer-based methods for accurate detection and classification of bacteria species significantly reduce the cost, time, and avoiding scientists the risk of infection. This paper presents a novel computer-based approach for highly accurate recognition of bacterial species. The proposed method consists of two main stages. First, a novel set of fractional-order orthogonal moments proposed to extract the fine features from the color images of bacteria. Second, a new method for feature selection, SSATLBO, is proposed. In this method, the teaching-based learning optimization (TLBO) as local operators is used to improve the exploitation ability of the Salp Swarm Algorithm (SSA) to avoid the local point. The proposed detection and classification method tested by using the DIBaS dataset (Digital Image of Bacterial Species), which includes 660 images with 33 various genera and classes of bacteria. The proposed method achieved a bacterial species recognition rate, 98.68%. The obtained results ensure the superiority of the proposed method over the traditional SSA and TLBO methods and the other Metaheuristic methods.