The Liancheng White Duck is a nationally protected breed in China, but its high-density farming environment poses significant challenges for target detection and behavior recognition, particularly due to occlusion, motion blur, and flock aggregation, making practical flock monitoring and counting labor intensive and prone to error in real barns. To address these issues, we propose DenseDuckMOT, an integrated detection-tracking framework for practical farm monitoring using existing fixed surveillance cameras with minimal additional hardware cost that combines the improved DuckNet detector with the AKFTrack tracker. DuckNet, based on YOLOv11, incorporates BiFPN, GLSA, and ESDH. It achieves high performance with 98.19% precision, 94.79% mAP@0.75, 97.70% F1-score, and 97.72% recall, while maintaining a lightweight design of only 1.90M parameters and a model size of 4485 KB. AKFTrack introduces adaptive Kalman prediction and a two-stage association scheme. It is evaluated on five dense white duck surveillance videos, where it outperforms or ranks second in MOTA, IDF1, and recall compared to DeepSORT, StrongSORT, and ByteTrack, especially in crowded and occluded scenes. Experimental results, ablation studies, and LayerCAM visualizations confirm the complementary advantages of BiFPN, GLSA, and ESDH, as well as the robustness of AKFTrack in handling occlusion and rapid motion. DenseDuckMOT provides accurate, efficient, and stable real-time monitoring in dynamic poultry farms, offering a scalable solution for intelligent farming.

Cat behavior recognition in unconstrained videos is important for animal welfare monitoring and veterinary assessment, yet remains challenging because behavior cues are often carried by highly deformable and intermittently visible parts such as the head and tail. This study aims to improve clip-level cat behavior recognition under unstable part visibility in real-world videos. We propose PMTNet, a part-centric temporal network for cat behavior recognition under unstable part visibility. The framework first detects the cat body, head, and tail using a DEIM-based detector, then selects a detector according to video-domain continuity and stability, and finally models behavior from ROI appearance features and explicit geometric motion cues. The framework was developed and evaluated using a part-detection dataset of 4000 training images and 500 validation images, together with a cat behavior dataset of 1283 video clips across five categories. In the best-performing setting, PMTNet achieved 93.1% Top-1 Accuracy and 90.9% Macro-F1. Ablation studies further suggest that detector choice in the video domain, complementary part cues, and missing-aware fusion all contribute to the final recognition performance. On the present dataset, PMTNet also outperformed representative end-to-end video recognition baselines. These results support the use of part-centric temporal modeling for cat behavior recognition in unconstrained real-world videos.