Coyote Optimization Algorithm (COA) has demonstrated efficient performance by utilizing the multiple pack (subpopulation) mechanism. However, the fixed number of packs and a relatively singular evolutionary strategy limit its comprehensive optimization performance. Thus, this paper proposes a COA variant, referred to as the Optimization State-based Coyote Optimization Algorithm (OSCOA). In the OSCOA algorithm, a Population Optimization State Estimation Mechanism is employed for estimating the current population optimization state. Then, the estimation result is used to guide the algorithm in setting the number of packs appropriately as well as selecting appropriate evolutionary strategies to refine search directions, thereby avoiding blind exploration. Additionally, the estimation result assists each pack in selecting suitable parents to generate pups, further improving the global search efficiency of the algorithm. To validate the effectiveness of the proposed algorithm, the OSCOA algorithm is subjected to comprehensive testing and analysis along with seven efficient optimizers on 71 benchmark functions derived from the CEC2014, CEC2017, and CEC2022 benchmark suites. The results of these extensive experiments indicate the competitive performance of OSCOA. Furthermore, to further assess the capability of the OSCOA algorithm in addressing real-world problems, two practical applications is considered: wireless sensor network deployment and image segmentation. The outcomes of these applications further confirm the efficacy and stability of the OSCOA algorithm in tackling real-world scenarios.