Real-time livestock identity recognition method

Abstract

The invention discloses a real-time livestock identity recognition method, and belongs to the technical field of precise animal husbandry, and the method specifically comprises the following steps: obtaining a livestock target detection data set through pasture video collection; constructing a livestock identification target detection network; obtaining a trained livestock identification target detection network; obtaining a predicted livestock face image; cutting the predicted livestock face image according to a livestock face frame in the predicted livestock face image, inputting the cut predicted livestock face image into the embedding and excitation network, and obtaining a predicted livestock feature vector through a linear regression function; matching the predicted livestock feature vector with an original livestock feature vector in a livestock feature database, and calculating the similarity of the feature vectors to obtain a livestock image with the highest similarity in the livestock feature database so as to complete livestock identity recognition; according to the scheme, the problems of uncertainty of manual feature extraction and limitation of electronic ear identification in an existing livestock identity identification method are solved.

Description

technical field [0001] The invention belongs to the technical field of livestock identification in intelligent animal husbandry, and in particular relates to a real-time livestock identification method. Background technique [0002] At present, the traditional animal husbandry is constantly looking for a new development model, no longer focusing on increasing production, but paying more attention to the quality and variety of products, developing in the direction of intelligence, refinement, and scale, and put forward the development concept of precision animal husbandry , and apply more modern technologies and theories to the development of traditional animal husbandry, use scientific concepts to guide the various links of animal husbandry from breeding production to residents' consumption, and improve the production efficiency of each link. In the breeding process of animal husbandry, livestock, as a single individual with high economic value, has extremely high research v...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the above-mentioned deficiencies in the prior art, a real-time livestock identification method provided by the present invention solves the problems of uncertainty of manually extracted features and limitations of electronic earmark identification in existing livestock identification methods

Images