Fat Content Detection Of B-scan Images Of Pigs Based On Deep Learning

The detection of pork fat content is of great significance to the livestock traders of pig breeding,and is also essential to the Chinese people who have higher and higher requirements on the quality of pork.Compared with various methods used to detect the contents of intramuscular fat(IMF)in pigs,the traditional methods of manual or physicochemical detection after slaughtering and raising pigs are highly destructive and require high technical requirements of employees,which cannot achieve nondestructive testing.In recent years,deep learning,as a branch of machine learning,has become a research hotspot in the field of computer vision.In particular,convolutional neural network,as one of the main research algorithms of deep learning,has made a breakthrough in object detection and image processing tasks.This paper provides an idea for the detection of fat content in pig flesh by using deep learning algorithm and realizes the nondestructive detection of fat content by analyzing the moving b-ultrasound image of pork.In this paper,135 groups of b-ultrasound images and physicochemical test data of porcine eye muscle were taken as experimental samples,and deep learning technology was used to study the nondestructive testing method of fat content detection in b-ultrasound images of porcine eye muscle.The objective of the study is to establish a convolution network prediction model to predict the fat content in the porcine eye muscle region from b-ultrasound images.Based on b-ultrasound image and fat content label of porcine eye muscle,a predictive regression model was established to detect fat content in b-ultrasound image of porcine eye muscle.Forecasting and then,in view of low efficiency and poor space continuity problems,aiming at the shortage of the convolutional neural network prediction model,the paper also put forward based on the depth of the cascade network pig fat ultrasound image detection method,this paper based on the convolutional neural network will first pig eye muscle area partition,this method can quickly from the pig in the ultrasound image segmentation in pig eye muscle area,then the pig eye muscle fat content prediction method,this method can not only effectively segment the complete pig eye muscle area,also significantly improved the pig eye muscle fat content prediction accuracy.The main contents of the paper are as follows:(1)data sample pretreatment.Due to the problem of insufficient image quality and quantity of the data set provided in this paper,image enhancement and data enhancementshould be carried out on the data set before training the model,and the b-ultrasound image of pig eye muscle should be labeled according to the needs of the model.(2)to study a method of fat content detection in b-ultrasound images of porcine eye muscle based on deep learning,and construct a regression prediction model using convolutional neural network.LeNet is mainly used for handwritten character recognition and has been put into use in American Banks.This indicates that the modified model can effectively identify pig b-ultrasound images and predict the fat content.But there are improvements in accuracy.(3)in view of the d eficiency of pig b-super fat content detection based on convolutional neural network,the paper introduces full convolutional network and proposes a method of pig b-super fat content detection based on deep cascade network.The method is divided into two stages: the first stage trains the full convolutional network to locate and segment the porcine eye muscle,and the second stage trains the deep network to predict the porcine fat content.The experimental results show that this method not only improves the accuracy of detecting pig fat content,but also improves the prediction efficiency.(4)according to the improved method for detecting b-super fat content of porcine eye muscle,develop a set of b-super image fat content detection system with high accuracy and strong robustness for b-super manufacturers.