Method for individual cow identification

Abstract

The invention relates to a method for individual cow identification, and relates to an image identification method in image data processing. The method comprises the following steps: collecting cow video data; converting the cow video data into an image sequence, and normalizing the image sequence; extracting four different component features of individual cows through wavelet transform; reducing the dimension of the four different component features through an improved KPCA method; carrying out weighting fusion on the four different component features; dividing image feature data into a training set and a test set; carrying out SVM training and SVM testing; and finally, identifying individual cows. The method overcomes the problem in the prior art that the intelligent identification method for individual cows still adopts a general image processing method instead of a suitable image processing method selected according to the individual characteristics of cows and especially the unique black-and-white patterns of cows, which results in low accuracy of individual cow identification.

Description

technical field [0001] The technical solution of the invention relates to an image recognition method in image data processing, in particular to a method for individual recognition of dairy cows. Background technique [0002] With the rapid development of information technology and the accelerated pace of agricultural modernization, large-scale and information-based dairy farms have been widely used. my country has basically formed a high-density and centralized dairy farming system. Dairy cow individual identification, as the basis of dairy cow information management, has become a research hotspot in the development of animal husbandry. With the continuous deepening of research in the fields of artificial intelligence and pattern recognition, image processing is developing in the direction of high speed, high resolution, three-dimensional, multimedia, intelligence and standardization. Using image processing methods to link dairy cow individual recognition with computer vis...

AI Technical Summary

Problems solved by technology

However, this method is highly dependent on the environment in which individual cows live. If they are in a natural environment with light or complex backgrounds, the recognition accuracy is difficult to guarantee.

In the paper "Muzzle-Based cattle identification using speed uprobust feature approach" published by the IEEE Computer Society, Ahmed S et al. proposed to use the accelerated robust feature (SURF) operator to extract the cow mouth texture feature points, and then use the SVM classifier A method for classifying and identifying individual dairy cows, but this method is sensitive to the location of the collected data, and it is difficult to obtain the data of the cow's mouth

CN105260750A discloses a cow individual recognition method. The method matches the captured real-time image of the cow with the pre-established template image library, and takes the label of the cow individual in the template library that can be successfully matched as the category of the cow to be identified , but this method is less efficient, and the number of templates that need to be traversed for each matching process may be large

Cai C et al. proposed a local binary pattern (LBP) texture feature method based on the cow face model in the paper "Cattle face recognition using local binary pattern descriptor" published at the Signal and Information Processing Association Summit, through sparse The optimization and low-rank decomposition are used to improve the robustness of the image processing method to effectively overcome the influence of light, image dislocation and occlusion in the image. The results show that LBP can correctly identify individual cows to a certain extent, but the face of the cow is relatively large compared to the cow. The torso contains less individual-specific information, and the recognition accuracy is still not ideal

CN201710000628.5 discloses a cow individual recognition method based on a deep convolutional neural network. The model parameter setting of this method requires high skill and requires repeated experiments for training and testing. There is no unified reference standard, and the establishment of the model is relatively difficult. difficulty

[0004] In the prior art, most of the intelligent identification of individual dairy cows adopts the traditional general-purpose image processing method, which does not target the individual characteristics of dairy cows, especially does not select a suitable image processing method based on the uniqueness of the black and white pattern shape of the cow's trunk, resulting in the failure of individual identification of dairy cows. low accuracy

Images