The basic behaviors of dairy cows (drinking, ruminating, walking, standing and lying) are closely related to their physiological health status. Consequently, intelligent behavior recognition is of significance for the automatic diagnosis and precision farming of dairy cows. Realizing the accurate behaviors classification in complex environments involving low quality surveillance videos, complex illumination and weather changes is a key problem in dairy farming that must be solved. In this study, CNN-LSTM (fusion of convolutional neural network and long short-term memory) an algorithm for recognizing the basic behaviors of a single cow, was proposed. First, the VGG16 trained on ImageNet was used as the network skeleton to extract the feature vector sequence corresponding to each video, so as to avoid the shortcomings of traditional feature engineering which were time-consuming and laborious. Then, these features were input Bi-LSTM (bidirectional long short-term memory) classification model, which could extract semantic information of time series data in two directions, so as to realize accurate recognition of dairy cow’s basic behaviors. To verify the effectiveness of the VGG16 feature extraction network used in this research, 1370 segments of approximately 18 h of videos collected from dairy farm monitoring cameras were tested and compared with those of five different feature extraction networks based on VGG19, ResNet18, ResNet101, MobileNet V2 and DenseNet201. Moreover, the effects of changes in illumination, weather, and wind velocity on behaviors recognition were tested and discussed. The test results indicated that the precision of the proposed algorithm for the recognition of the five behaviors ranged from 0.958 to 0.995, the recall ranged from 0.950 to 0.985, and the specificity ranged from 0.974 to 0.991, while the average precision, recall and specificity were 0.971, 0.965 and 0.983, respectively. The average recognition accuracy of the proposed method was 0.976, which was higher than the methods based on VGG19, ResNet18, ResNet101, MobileNet V2 and DenseNe201 by 0.08 × 10⁻², 1.97 × 10⁻², 2.19 × 10⁻², 2.85 × 10⁻² and 2.34 × 10⁻², respectively. Furthermore, the influences of illumination, weather and wind speed on the algorithm were discussed. The results showed that the difference of behavior recognition accuracy of this method under the above interference was less than 0.02, indicating that the method is good in stability. The research results showed that it was feasible to use the proposed algorithm to recognize behaviors of a single target dairy cow. This study could not only provide valuable references for the behaviors identification and understanding of multiple target dairy cows based on computer vision in complex environments such as low-quality surveillance video, complex illumination and weather variations, but also contribute to their physiological health assessment and remote diagnosis. The study may be valuable for the dairy cows’ prevention and treatment of health and reproduction problems using the “medical-engineering interdisciplinary” approach.

奶牛的基本行为（饮水，反刍，行走，站立和躺卧）与其生理健康状况密切相关。因此，智能行为识别对于奶牛的自动诊断和精确饲养具有重要意义。在涉及低质量监控视频，复杂照明和天气变化的复杂环境中实现准确的行为分类是奶业必须解决的关键问题。在这项研究中，提出了CNN-LSTM（卷积神经网络和长短期记忆的融合）一种识别单头母牛基本行为的算法。首先，将在ImageNet上训练的VGG16用作网络骨架，以提取与每个视频相对应的特征向量序列，从而避免了传统特征工程的费时费力的缺点。然后，将这些特征输入到双向长期短期记忆（Bi-LSTM）分类模型中，该模型可以从两个方向提取时间序列数据的语义信息，从而实现对奶牛基本行为的准确识别。为了验证本研究中使用的VGG16特征提取网络的有效性，测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet的五个不同特征提取网络的视频进行了比较V2和DenseNet201。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 这些特征通过输入双向LSTM（双向长短期记忆）分类模型，可以从两个方向提取时间序列数据的语义信息，从而实现对奶牛基本行为的准确识别。为了验证本研究中使用的VGG16特征提取网络的有效性，测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet的五个不同特征提取网络的视频进行了比较V2和DenseNet201。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 这些特征被输入双向双向LSTM（双向长期短期记忆）分类模型，可以从两个方向提取时间序列数据的语义信息，从而实现对奶牛基本行为的准确识别。为了验证本研究中使用的VGG16特征提取网络的有效性，测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet的五个不同特征提取网络的视频进行了比较V2和DenseNet201。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 它可以从两个方向提取时间序列数据的语义信息，从而实现对奶牛基本行为的准确识别。为了验证本研究中使用的VGG16特征提取网络的有效性，测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet的五个不同特征提取网络的视频进行了比较V2和DenseNet201。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 它可以从两个方向提取时间序列数据的语义信息，从而实现对奶牛基本行为的准确识别。为了验证本研究中使用的VGG16特征提取网络的有效性，测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet的五个不同特征提取网络的视频进行了比较V2和DenseNet201。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet V2和DenseNet201的五个不同特征提取网络的视频进行了比较。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明 测试了从奶牛场监控摄像机收集的大约18小时视频中的1370段，并将其与基于VGG19，ResNet18，ResNet101，MobileNet V2和DenseNet201的五个不同特征提取网络的视频进行了比较。此外，测试和讨论了光照，天气和风速变化对行为识别的影响。测试结果表明该算法对五种行为的识别精度为0.958至0.995，召回范围为0.950至0.985，特异性为0.974至0.991，平均精度，召回度和特异性分别为0.971、0.965和0.983，分别。平均识别精度所提出的方法是0.976，这是由0.08×10比基于VGG19，ResNet18，ResNet101，MobileNet V2和DenseNe201方法更高^-2，1.97×10 ^-2，2.19×10 ^-2，2.85× 10 ^-2和2.34×10 ^-2， 分别。此外，讨论了光照，天气和风速对算法的影响。结果表明，该方法在上述干扰下的行为识别精度差异小于0.02，说明该方法具有良好的稳定性。研究结果表明，使用该算法识别单个目标奶牛的行为是可行的。这项研究不仅可以为复杂环境下基于计算机视觉的多目标奶牛的行为识别和理解提供有价值的参考，例如低质量的监控视频，复杂的光照和天气变化，还有助于它们的生理健康评估和远程诊断。