Optimizing Muzzle Pattern Identification through Improved Pattern Recognition Techniques

The global pet industry and market have been thriving in recent years, and the rise in pet population has led many countries to implement policies and systems for their management. The South Korean government has also introduced an animal registration system under the Animal Protection Act, but only about 37% of the country's pet population are registered in the system due to the inconvenient methods of registration. Biometric technology, in general, is used to identify individuals based on accurate object recognition. This paper applies Boosted Efficient Binary Local Image Descriptor (BEBLID) to the commonly-used ORB algorithm with the goal to improve the recognition and matching of muzzle pattern data and to derive the optimal value of the BEBLID scale coefficient (K) for muzzle pattern recognition. A total of 200 muzzle patterns were collected from dogs to use as the data for analysis. To demonstrate the superiority of the proposed method, the ORB algorithm was used as a benchmark. The matching rate achieved when BEBLID's K value was set to the default value of 1 was 76.24%, compared to the 66.82% matching rate of the ORB algorithm. Also, the optimal K was determined to be 0.75, achieving an 87% matching rate, after testing with variant K values from 0.25 to 2. Overall, this study demonstrates that by applying BEBLID to traditional ORB algorithms using the optimal scale coefficient value, the recognition and matching rate can be significantly improved. The commercialization and practical application of the muzzle pattern recognition technique proposed in this study is expected to contribute to improving the pet registration rate, which has been stagnant despite its necessity, and the management of pet populations.