针对白羽肉鸡体质量测量自动化水平低、易造成肉鸡应激的问题,提出一种结合深度学习的非接触式白羽肉鸡体质量估测方法。利用Mask R-CNN和YOLACT(You only look at coefficients)两种实例分割算法获取白羽肉鸡位置与覆盖掩膜,并进行效...针对白羽肉鸡体质量测量自动化水平低、易造成肉鸡应激的问题,提出一种结合深度学习的非接触式白羽肉鸡体质量估测方法。利用Mask R-CNN和YOLACT(You only look at coefficients)两种实例分割算法获取白羽肉鸡位置与覆盖掩膜,并进行效果对比;采用自适应掩膜随机提取白羽肉鸡身体部分边缘点,并作为观测点进行椭圆拟合,映射白羽肉鸡背部像素投影面积;通过双变量相关性分析验证白羽肉鸡背部投影面积与体质量间的显著相关性,根据白羽肉鸡背部投影面积与背部像素投影面积的线性比例关系,按照最小二乘原则建立白羽肉鸡背部像素投影面积与体质量间的线性回归模型。试验表明,单只鸡体质量估测中以Mask R-CNN进行特征提取的体质量估测平均准确率为97.23%,以YOLACT进行特征提取的体质量估测平均准确率为97.49%,群鸡场景中体质量估测最低准确率为90.50%。展开更多
Multiply scattered waves are sensitive to media changes owing to the effect of repeated sampling,superposition and amplification. Based on this characteristic,small-medium changes could be detected by using coda wave ...Multiply scattered waves are sensitive to media changes owing to the effect of repeated sampling,superposition and amplification. Based on this characteristic,small-medium changes could be detected by using coda wave interferometry. In recent years,coda wave interferometry has been widely used in estimating velocity variation with high precision in areas such as seismology and non-destructive testing. This paper systematically presents the principle and research status of coda wave interferometry,and especial focus is placed on the research of media velocity variations by using repeating earthquakes,artificial sources,and ambient noise. Applications of coda wave interferometry can contribute to the more subtle understanding of dynamic evolution process in the medium.展开更多
文摘针对白羽肉鸡体质量测量自动化水平低、易造成肉鸡应激的问题,提出一种结合深度学习的非接触式白羽肉鸡体质量估测方法。利用Mask R-CNN和YOLACT(You only look at coefficients)两种实例分割算法获取白羽肉鸡位置与覆盖掩膜,并进行效果对比;采用自适应掩膜随机提取白羽肉鸡身体部分边缘点,并作为观测点进行椭圆拟合,映射白羽肉鸡背部像素投影面积;通过双变量相关性分析验证白羽肉鸡背部投影面积与体质量间的显著相关性,根据白羽肉鸡背部投影面积与背部像素投影面积的线性比例关系,按照最小二乘原则建立白羽肉鸡背部像素投影面积与体质量间的线性回归模型。试验表明,单只鸡体质量估测中以Mask R-CNN进行特征提取的体质量估测平均准确率为97.23%,以YOLACT进行特征提取的体质量估测平均准确率为97.49%,群鸡场景中体质量估测最低准确率为90.50%。
基金sponsored by the Task of Department of Earthquake Monitoring and Prediction,CEAthe Special Fund of Earthquake Scientific Research of China Earthquake Administration(201208003)
文摘Multiply scattered waves are sensitive to media changes owing to the effect of repeated sampling,superposition and amplification. Based on this characteristic,small-medium changes could be detected by using coda wave interferometry. In recent years,coda wave interferometry has been widely used in estimating velocity variation with high precision in areas such as seismology and non-destructive testing. This paper systematically presents the principle and research status of coda wave interferometry,and especial focus is placed on the research of media velocity variations by using repeating earthquakes,artificial sources,and ambient noise. Applications of coda wave interferometry can contribute to the more subtle understanding of dynamic evolution process in the medium.