<p>With the rapid development of the automotive industry and the national economy, the number of consumers purchasing cars has been increasing. This trend has also intensified the challenges faced by traffic management departments in regulating traffic. A license plate is the most important identifier of a vehicle, and how to efficiently recognize license plates has become a hot research topic. The key step in license plate recognition is edge detection, and currently, edge detection based on intelligent optimization algorithms has significant advantages. In this paper, we proposed a complete license plate recognition model, in which the core of the edge detection module is our proposed algorithm that ant colony optimization using eight- neighborhood search, with dual pheromone update strategy and continuous edge generation strategy (ACOEDC). This algorithm enhances edge connectivity and adaptability to complex structures through an eight-neighborhood search strategy, accelerates convergence speed and improves global search capability through a dual pheromone update strategy, and effectively improves the smoothness and integrity of edge curves through a continuous edge generation strategy, thereby achieving more accurate and stable edge detection results. Experiments on 12 different license plate samples with three difficulty levels showed that ACOEDC outperformed six intelligent optimization algorithms (FOGPSO, SAOk-AUS, IDO, XMACO, SFACO, and RESACO) in terms of accuracy and stability. The Wilcoxon rank sum test further confirmed the statistical significance of these results, demonstrating that ACOEDC is a novel algorithm. Based on the research questions (RQs), we analyzed the performance comparison between ACOEDC and these six intelligent optimization algorithms. The analysis results indicated that ACOEDC achieved the highest recognition accuracy and significant stability in license plate images of various difficulties, demonstrating strong robustness and generalization ability.</p>