This research investigates the efficacy of weed recognition models in cotton fields through advanced imaging and machine learning techniques. Utilizing 10 trials, the models, namely K-NN and GBM, were evaluated across multiple performance metrics. Results reveal that GBM consistently outperformed K-NN in accuracy, precision, recall, and F1 score, with average values of 0.88, 0.89, 0.86, and 0.88, respectively, compared to K-NN's averages of 0.85, 0.87, 0.82, and 0.85. Moreover, GBM exhibited higher AUC values (0.94) than K-NN (0.92) in ROC curve analysis, indicating superior discrimination ability. Additionally, k-fold cross-validation demonstrated GBM's higher mean accuracy (0.89) and F1 score (0.88) compared to K-NN (mean accuracy: 0.86, mean F1 score: 0.85). Additionally, integrating temporal data analysis could improve the models’ ability to detect weed growth patterns over time. Real-time monitoring capabilities and automated decision-making systems could streamline weed management practices in agricultural settings. Furthermore, expanding the study to encompass diverse geographical regions and crop types would provide valuable insights into the generalizability and robustness of the developed models. Overall, continued research in this domain holds the potential to revolutionize weed management strategies and contribute to sustainable agriculture practices.