| 基于生成对抗网络的禽蛋图像数据生成研究 |
| |
| 引用本文： | 李庆旭,王巧华,马美湖.基于生成对抗网络的禽蛋图像数据生成研究[J].农业机械学报,2021,52(2):236-245. |
| |
| 作者姓名： | 李庆旭 王巧华 马美湖 |
| |
| 作者单位： | 华中农业大学;华中农业大学；农业农村部长江中下游农业装备重点实验室;国家蛋品加工技术研发分中心 |
| |
| 基金项目： | 国家自然科学基金面上项目（31871863） |
| |
| 摘 要： | 在进行禽蛋无损检测研究时,需要花费大量的人力和物力采集禽蛋图像数据,为解决该问题,设计了一种基于深度卷积生成对抗网络(Deep convolutional generative adversarial networks,DCGAN)的改进禽蛋图像数据生成网络。该网络分为生成器与判别器,生成器用于禽蛋图像数据生成,判别器对生成的禽蛋图像进行真实性判断,两者相互对抗最终生成高质量的禽蛋图像数据。为了提高生成的禽蛋图像质量,使用残差网络构建生成器和判别器,引入Wasserstein距离和加梯度惩罚的损失函数,分别在透射和反射情况下对禽蛋图像进行生成研究。该方法有效地解决了大量禽蛋图像数据的采集问题,为后期禽蛋图像识别与检测提供了数据基础,同时也为后续禽蛋数据库构建提供了技术支持。
|
| 关 键 词： | 禽蛋图像 数据生成 生成对抗网络 Wasserstein距离 残差网络 |
| 收稿时间： | 2020/4/22 0:00:00 |
Poultry Egg Image Data Generating Based on Generative Adversarial Network |
| |
| LI Qingxu,WANG Qiaohu,MA Meihu.Poultry Egg Image Data Generating Based on Generative Adversarial Network[J].Transactions of the Chinese Society of Agricultural Machinery,2021,52(2):236-245. |
| |
| Authors: | LI Qingxu WANG Qiaohu MA Meihu |
| |
| Institution: | (College of Engineering,Huazhong Agricultural University,Wuhan 430070,China;Key Laboratory of Agricultural Equipment in Mid-lower Yangtze River,Ministry of Agriculture and Rural Affairs,Wuhan 430070,China;National Egg Processing Technology Research and Development Sub Center,Wuhan 430070,China) |
| |
| Abstract: | China is a big country in the production and consumption of poultry eggs, and testing the quality of poultry egg is very important. Machine vision plays an important role in non-destructive testing of poultry eggs. Computer vision is one of the commonly used technical methods in the field of non-destructive testing of poultry eggs. This technology needs to collect a large number of poultry egg image data for analysis and modeling in order to obtain a good recognition effect. When doing research on non-destructive testing of poultry eggs, researchers need to spend a lot of manpower and material resources on the collection of image data of poultry eggs. Aiming to solve this problem, an improved egg image data generation network based on unsupervised representation learning with deep convolutional generative adversarial networks (DCGAN) was proposed. The network was divided into a generator and a discriminator. The generator was used to generate the egg image data. The discriminator judged the authenticity of the generated egg image. The generator and discriminator competed with each other and finally generated high-quality egg image data. In order to improve the quality of the generated egg images, a residual network was used to construct a generator and discriminator, the loss functions of Wasserstein distance and gradient penalty were introduced to research the image generation in the case of egg transmission and egg reflection. This method effectively solved the problem that it was difficult to collect a large number of poultry egg image data, provided a data basis for the later identification and detection of poultry egg image, and also provided technical support for the subsequent establishment of poultry egg database. |
| |
| Keywords: | poultry egg image data generating generative adversarial network Wasserstein distance residual network |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《农业机械学报》浏览原始摘要信息 |
| 点击此处可从《农业机械学报》下载免费的PDF全文 |
|
