Recognition and localization of maize seedlings based on computer vision (Retraction Notice)

To address the issues of inaccurate crop recognition and poor localization precision in cornfield weeding robots, this study proposes a precise identification method that integrates computer vision techniques. The approach begins with background segmentation to isolate plants from the soil, employs morphological erosion to remove noise and highlight morphological features, and finally utilizes the thinning algorithm to extract the maize crop seedings and determine their center points. Tests on 100 seedling-stage images demonstrated that the localization error for corn stalk centers is less than 12 mm. This research provides a reliable technical solution for enabling precise and non-damaging autonomous operation of weeding robots.