Cow face identification method based on convolutional neural network and classifier model

Abstract

The invention belongs to the computer vision and intelligent identification technology field and especially relates to a cow face identification method based on a convolutional neural network and a classifier model. A last hidden layer of the convolutional neural network is a fully-connected layer containing 32, 64, 128, 256 or 512 nerve cells and is used for extracting a characteristic. And then, the classifier model is used to complete identification of a cow individual. When there is a newly-added cow, image data of the cow only needs to be collected, the data is input into a convolutional neural network model, and a characteristic is extracted and is added to an original classification model so that identification can be performed; and the convolutional neural network model does not need to be retrained. In the invention, the convolutional neural network model of a 64-dimension characteristic extraction layer is selected, sparsity is used to express a classification model, training data 24000 pictures and test data 6000 pictures of 30 cows, which are randomly selected, are tested, and a result shows that identification time is shortened through using the method; average time consuming for identifying each cow is shortened to 0.00022s; and an identification rate reaches more than 99%.

Description

technical field [0001] The invention belongs to the technical field of computer vision and intelligent recognition, and relates to a cow individual recognition technology, in particular to a cow face recognition method based on a convolutional neural network and a sparse representation classification model. Background technique [0002] With the improvement of our national living standards, people's demand for beef products and milk products is increasing day by day, and at the same time, their attention to their quality is also increasing. It has become an urgent need to increase the total output and quality of beef and milk. The development of the breeding industry must realize the intelligence, scale, automation and standardization of cattle breeding. Therefore, digital and refined breeding based on the individual body condition of cattle has become the main development direction of modern scientific cattle breeding. [0003] As the basis of intelligent breeding managemen...

AI Technical Summary

Problems solved by technology

However, the use of RFID electronic ear tags for individual cattle identification has the following problems: First, the application cost of the production and circulation stages is high, resulting in the high price of RFID electronic ear tags, which is difficult to use on a large scale; The housing environment is quite different, the relevant RFID standards are not uniform, and the RFID recognition distance, recognition accuracy and other technical differences are relatively large; the third is that the electronic ear tags need to be installed on the cattle, which brings harm and pain to the cattle ; Fourth, there are interference sources in the barn, and the electronic ear tags are easy to fall off or be replaced by others, resulting in a decrease in the reliability of the identification system

Second, when the image is collected, the position of the cow target in the image is random, and the posture is not fixed. The recognition algorithm must overcome changes in light and shade, displacement, affine, etc.

CN106778902A discloses a cow individual recognition method based on a deep convolutional neural network, which adopts deep learning to extract features from the convolutional neural network, and combines the texture features of the cow's back pattern to effectively identify the individual cow; the steps of the method are: Collection of dairy cow data, preprocessing of training set and test set, design of convolutional neural network, training of convolutional neural network, generation of recognition model, identification of individual cows using the recognition model; the method uses optical flow method or frame difference method Extract cow torso images, use convolutional neural network to extract features, and combine texture features of cows to realize effective identification of cows, but this method is only suitable for a type of cows with obvious back flower patterns, and when the number of cows increases Retrain the neural network. At the same time, due to the large number of neurons in each layer of the neural network, the number of network parameters is too large, which is prone to overfitting and reduces the generalization ability of the network.

[0007] To sum up, the main problems existing in the existing cattle identification technology are: 1. RFID electronic ear tag identification technology: the identification range is limited, and electronic tags need to be installed on the cattle, which brings harm and pain to the cattle, and Electronic tags are easy to fall off or be replaced by people, which reduces the accuracy and reliability of recognition; 2. Traditional image feature extraction and classification methods: the recognition rate is low in complex environments

The background of the cattle pictures collected in the farm environment is messy, there are occlusions, the lighting conditions change greatly, and the posture of the cattle is not fixed. The probability of wrong matching of the key point information extracted by the traditional image feature extraction method increases, and the accuracy Not high, when the amount of data is large, it will take a long time, and real-time accurate identification cannot be achieved; 3. Convolutional neural network technology: only use this technology to identify individual cattle, and when the number of cattle increases, the convolutional neural network needs to be retrained A neural network, which takes a long time and the neural network cannot be used during training

Images