Design and optimization of a bird’s nest foreign object removal robot

To enhance the automation level of foreign object removal in bird’s nest processing and reduce the costs and misjudgment rates associated with manual inspection, this paper proposes and develops a parallel system for automatic foreign object removal in bird’s nest based on machine vision. A two-stage foreign object detection algorithm is proposed: the first stage introduces an attention mechanism to improve the U-Net++ model, enhancing segmentation accuracy under texture interference; the second stage combines binary mask processing to achieve precise classification of foreign object and non-foreign object regions. Experimental results show that the algorithm achieves an F 1 score of 94.80% and a recall rate of 97.90%, outperforming traditional methods in overall performance. An evaluation system for bird’s nest foreign object removal is established, incorporating criteria such as cleanliness, loss rate, structural integrity, and removal time. A set of automated equipment with identification, localization, and removal functions is designed and implemented. The algorithm and equipment were deeply integrated, and through prototype construction and performance testing, the system’s ability to accurately identify and stably remove various types of foreign objects under complex operating conditions was validated. This study provides a feasible intelligent solution for the bird’s nest impurity removal field, with good application prospects and industrial promotion value.