叶长文,戚超,刘超,郑小刚,王鹏,陈坤杰.基于Faster-RCNN的肉鸡击晕状态检测方法[J].农业机械学报,2019,50(12):255-259.
YE Changwen,QI Chao,LIU Chao,ZHENG Xiaogang,WANG Peng,CHEN Kunjie.Stunning State Recognition Method of Broiler Chickens Based on Faster Region Convolutional Neural Network[J].Transactions of the Chinese Society for Agricultural Machinery,2019,50(12):255-259.
摘要点击次数: 1752
全文下载次数: 656
基于Faster-RCNN的肉鸡击晕状态检测方法   [下载全文]
Stunning State Recognition Method of Broiler Chickens Based on Faster Region Convolutional Neural Network   [Download Pdf][in English]
投稿时间:2019-04-17  
DOI:10.6041/j.issn.1000-1298.2019.12.029
中文关键词:  肉鸡  电击晕  击晕状态  卷积神经网络  深度学习
基金项目:“十二五”国家科技支撑计划项目(2015BAD19806)和国家肉鸡产业技术体系项目(CARS-42-5)
作者单位
叶长文 南京农业大学 
戚超 南京农业大学 
刘超 南京农业大学 
郑小刚 江苏省农机化服务站 
王鹏 南京农业大学 
陈坤杰 南京农业大学 
中文摘要:为了准确识别屠宰加工中肉鸡的击晕状态,提出了一种基于快速区域卷积神经网络的肉鸡击晕状态检测方法。对输入图像进行归一化处理,通过卷积神经网络(VGG16)提取肉鸡的卷积特征图,利用区域建议网络提取预测框,在卷积特征图上采用非极大值抑制算法去除重复表述的预测框;将所得的各预测框映射到卷积特征图上,得到预测框在卷积特征图上的候选区域,将其输入感兴趣区域池化层;通过感兴趣区域池化层将大小不一的候选区域进行池化操作、得到统一的输出数据,最后通过全连接层与柔性最大值分类器,输出各击晕类别的概率和预测框的坐标。将2319个样本图像按2∶1的比例随机分为训练集与测试集,对模型进行训练与实验验证。结果表明,本文建立的基于Faster-RCNN的肉鸡击晕状态分类模型对773个测试集肉鸡样本击晕状态分类的总准确率达到96.51%,对肉鸡击晕状态的预测速度可达每小时37000只,基本满足肉鸡屠宰生产线要求。
YE Changwen  QI Chao  LIU Chao  ZHENG Xiaogang  WANG Peng  CHEN Kunjie
Nanjing Agricultural University,Nanjing Agricultural University,Nanjing Agricultural University,Agricultural Mechanization Service Station of Jiangsu Province,Nanjing Agricultural University and Nanjing Agricultural University
Key Words:broiler chickens  electric stunning  stunning state  convolutional neural network  deep learning
Abstract:In order to improve the accuracy of stunning state recognition of broiler chickens, a method of stunning state classification of broilers based on regional convolutional neural network (RCNN) was proposed. The following method was able to detect insufficiently appropriately and excessively stunned conditions of broilers. Initially, the image acquisition platform was utilized to collect the sample images. The data sets of collected samples were made according to the PASCAL visual object classes data set format. The total samples of 2319 images were randomly divided into training set and test set with the ratio of 6∶3. The augmented training sets were obtained through image enhancement technology. A Faster-〖JP〗RCNN was trained by using the augmented training set to detect the stunning states of broilers. The results showed that the recognition accuracy of the Faster-RCNN was 96.51% for 773 sample images in the test set. The accuracy of Faster-RCNN model was significantly higher than that of the established back propagation neural network (BP-NN) model (90.11%). The proposed model could be used to inspect the stunning state of more than 37000 broilers per hour. Deep learning technology was applied to recognize the stunning states of broilers, which can be used to automatically detect the stunning state of broilers and enhance automated slaughtering processes in the poultry industry. 

Transactions of the Chinese Society for Agriculture Machinery (CSAM), in charged of China Association for Science and Technology (CAST), sponsored by CSAM and Chinese Academy of Agricultural Mechanization Science(CAAMS), started publication in 1957. It is the earliest interdisciplinary journal in Chinese which combines agricultural and engineering. It always closely grasps the development direction of agriculture engineering disciplines and the published papers represent the highest academic level of agriculture engineering in China. Currently, nearly 8,000 papers have been already published. There are around 3,000 papers contributed to the journal each year, but only around 600 of them will be accepted. Transactions of CSAM focuses on a wide range of agricultural machinery, irrigation, electronics, robotics, agro-products engineering, biological energy, agricultural structures and environment and more. Subjects in Transactions of the CSAM have been embodied by many internationally well-known index systems, such as: EI Compendex, CA, CSA, etc.

   下载PDF阅读器