Chicken individual identification method based on VGG-16 convolutional neural network

Abstract

The invention provides a chicken individual identification method based on a VGG-16 convolutional neural network. The method includes collecting home chicken shadow image data by using a mobile cameraand intercepting a video corresponding to each individual into a picture by using Open CV according to a frame rate; changing the size of the picture into 100px * 100px by using a Python PIL toolkit,adding a prefix 1 _ to an output picture of the meadow chicken; adding a prefix 2 _ to an output picture of the white feather chicken; adding a prefix 3 _ to an output picture of the green-foot partridge chicken; 80% of the obtained domestic chicken individual pictures are used as a training set; taking 15% of the training set as a test set and 5% of the training set as a test set for subsequenttests, establishing a Keras convolutional neural network model, compiling a corresponding VGG convolutional neural network, adding the training set and the test set into the neural network, performingiterative training, and performing iteration for 10000 times to obtain a weight file; inputting a picture, and predicting the domestic chicken variety according to the weight file, so that the domestic chicken detection precision is effectively improved, and the method can accurately, efficiently and robustly detect the domestic chicken variety in a complex chicken farm environment.

Description

Technical field: [0001] The present invention relates to the field of informationized animal husbandry, and more specifically, relates to a chicken individual identification method based on a VGG-16 convolutional neural network. Background technique: [0002] China has a huge consumer market for poultry products, which requires a large scale and a large number of poultry farms. At present, the vast majority of farms are still using manpower for feeding and management. This manpower-dependent approach has adverse effects on both the chickens and the breeders themselves. [0003] For domestic chickens in chicken farms, the presence of humans will intimidate the chickens, affect the normal growth of domestic chickens, and ultimately affect the meat quality of broiler chickens and the egg production of laying hens. For chicken farm personnel, manual labeling is an unhealthy, tedious and error-prone work. When labeling the serial number and type of chicken, the breeder must ha...

AI Technical Summary

Problems solved by technology

For chicken farm personnel, manual labeling is unhealthy and tedious and error-prone

When labeling the serial number and type of chicken, the breeder must have close contact with the chicken cage and stay on the chicken farm for a long time, which has brought about the following adverse effects: First, the chickens carry more parasites, It is very easy to infect chicken farm personnel

Second, chicken farms contain a large amount of highly carcinogenic substances such as glutaraldehyde and formaldehyde, which will adversely affect the fertility of chicken farm personnel, and even cause changes

Third, because chicken farms are usually large-scale breeding, it takes a lot of manpower and time to mark each chicken manually, which will greatly increase the cost of breeding, and it is easy to cause wrong labeling and omissions due to human errors

Images