In breeding farm cattle feeding and watering scenarios, the number of cattle is dense, resulting in complex scenes and spatial congestion, and traditional single-modal cattle identification methods often encounter problems such as susceptibility to occlusion and low identification accuracy. Consequently, this study proposes a decision layer fusion cattle identity recognition method with multiple features of cattle face, cattle muzzle pattern, and cattle ear tag. The image is segmented into cattle face, cattle muzzle pattern, and cattle ear tag by the SOLO algorithm. Cattle face, and cattle muzzle patterns adopt the FaceNet network model with different backbones, respectively, while the ear tag adopts the PP-OCRv4 network model. The experiment extracts the features of the three parts respectively, stores these features in the database while comparing them with the database, calculates the Euclidean distance and text similarity with the sample data, and extracts the Top 3 confidence levels. Then, it performs One-Hot encoding processing for each of these features, which are used as inputs to the decision-making layer. An integration approach is used in the decision fusion part, where different integration strategies combine multiple base classifiers and compare the best performance as the final decision fusion recognition model. The results show that using the multimodal decision fusion method makes the recognition accuracy reach 95.74%, 1.4% higher than the traditional optimal unimodal recognition accuracy. The verification rate reaches 94.72%, 10.65% higher than the traditional optimal unimodal recognition verification rate. This fusion method achieves individual recognition and accurate management of cattle in breeding farms, bringing an efficient and accurate solution for the animal husbandry industry.