Using a CNN-LSTM for basic behaviors detection of a single dairy cow in a complex environment

Wu, Dihua; Wang, Yunfei; Han, Mengxuan; Song, Lei; Shang, Yuying; Zhang, Xinyi; Song, Huaibo

doi:10.1016/j.compag.2021.106016

Cited by 53 publications

(39 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Notably, the datasets collected contain 530,485 data rows at 25 Hz, i.e., the data are approximately 5.89 h long; such datasets tend to be tiny. By comparison, in [16], approximately 63 h in total were captured, and a CNN model with long short-term memory (CNN-LSTM) was used to detect five basic behaviors (drinking, ruminating, walking, standing, and lying). In [17], the active video was approximately 68 h long and applied to an LSTM model.…”

Section: A Challenges In Behavioral Datamentioning

confidence: 99%

Data Augmentation for Inertial Sensor Data in CNNs for Cattle Behavior Classification

Tokgoz

Fukawa

et al. 2021

IEEE Sens. Lett.

View full text Add to dashboard Cite

Cattle behavior monitoring is critical for understanding cattle welfare and health status. One of the most powerful and cost-effective monitoring methods is a neural network-based monitoring system that analyzes time series data from inertial sensors attached to cows. However, while deep learning has achieved many successes in pattern recognition, large-scale datasets are often required. When given a limited number of data, data augmentation is an extremely useful and low-cost preprocessing step for neural network-based systems. Data augmentation for inertial sensor data, however, has yet to be thoroughly investigated. This work proposes several inertial sensor data augmentation methods in a manner that fits the characteristics of cattle behavioral data. The proposed approaches are applied to the task of cattle behavior classification with convolutional neural networks, which is challenging given limited data. The classification performance increases from 83.07 to 94.43% with appropriate augmentation steps. In conclusion, the data augmentation approaches presented here can help to improve deep learning performance regarding cattle behavior classification and decrease the overall system cost stemming from data acquisition and labeling.

show abstract

Section: A Challenges In Behavioral Datamentioning

confidence: 99%

Data Augmentation for Inertial Sensor Data in CNNs for Cattle Behavior Classification

Tokgoz

Fukawa

et al. 2021

IEEE Sens. Lett.

View full text Add to dashboard Cite

show abstract

“…Yin et al [ 115 ] proposed the EfficientNet-LSTM model to extract spatial feature for the recognition of cows’ motion behaviours, which achieved 97.87% behaviour recognition accuracy in the antagonism of environmental robustness. Wu et al [ 13 ] proposed CNN-LSTM (a fusion of convolutional neural network and long short-term memory) for recognising the basic behaviours of a single cow. In the former work, the experimental results illustrated that the precision of the proposed algorithm for the recognition of five behaviours ranged from 0.958 to 0.995, that the recall ranged from 0.950 to 0.985, and that the specificity ranged from 0.974 to 0.991.…”

Section: Cattle Lameness Detection and Scoringmentioning

confidence: 99%

“…In large farms, group or interaction behaviours are also important for animal welfare and the corresponding management. Meanwhile, some tiny behaviours such as limping is part of basic walking behaviours, which is difficult to detect using a general network [ 13 ]. In terms of behavioural analysis, environmental conditions are prone to be ignored.…”

Section: Cattle Lameness Detection and Scoringmentioning

confidence: 99%

“…The timely detection of lameness is important for providing effective and inexpensive treatment and for preventing future ailments [ 12 ]. Meanwhile, cattle behaviour is an important indicator for animals’ health and welfare, which influence the quantity and quality of cattle products [ 13 ]. However, traditional lameness detection and behaviour recognition approaches are time-consuming and labour-intensive, resulting in major concerns on farms.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Intelligent Perception-Based Cattle Lameness Detection and Behaviour Recognition: A Review

Qiao

Kong

Clark

et al. 2021

Animals

View full text Add to dashboard Cite

The growing world population has increased the demand for animal-sourced protein. However, animal farming productivity is faced with challenges from traditional farming practices, socioeconomic status, and climate change. In recent years, smart sensors, big data, and deep learning have been applied to animal welfare measurement and livestock farming applications, including behaviour recognition and health monitoring. In order to facilitate research in this area, this review summarises and analyses some main techniques used in smart livestock farming, focusing on those related to cattle lameness detection and behaviour recognition. In this study, more than 100 relevant papers on cattle lameness detection and behaviour recognition have been evaluated and discussed. Based on a review and a comparison of recent technologies and methods, we anticipate that intelligent perception for cattle behaviour and welfare monitoring will develop towards standardisation, a larger scale, and intelligence, combined with Internet of things (IoT) and deep learning technologies. In addition, the key challenges and opportunities of future research are also highlighted and discussed.

show abstract

“…Artificial neural network, fuzzy logic classifier, and machine learning based approaches using animal physiology and climatic variables have been found promising in monitoring animal core behaviours and thermal status under experimental conditions (Sousa et al, 2016(Sousa et al, , 2018Tsai et al, 2020;Becker et al, 2021). Though the use of depth cameras and imaging can provide more accuracy, RGB images and videos of different qualities under all environmental conditions have been analysed with high accuracy (Wu et al, 2021). Considering the speed of technology development, these will very likely be useful under practical conditions in the near future.…”

Section: Depth Imaging Video Surveillance and Artificial Intelligencementioning

confidence: 99%

Automated Monitoring of Cattle Heat Stress and Its Mitigation

Islam

Lomax

Doughty³

et al. 2021

Front. Anim. Sci.

View full text Add to dashboard Cite

Climate change related global warming is likely to continue, despite all mitigation measures taken by humans, due to the lag effect of long-term anthropogenic activities. Warming of the atmosphere can impact worldwide cattle production directly by compromising health, welfare and productivity, and indirectly by reducing the quality and quantity of animal feed. Under warm thermal conditions, cattle adjust their physiological and behavioural responses as an integral part of thermoregulation to maintain internal body temperature within a safe range. However, a greater intensity and duration of heat exposure can exceed thermoregulatory capacity leading to an increase in internal body temperature beyond the normal limit that ultimately evokes different animal responses to heat. In cattle, response to heat stress can be visually observed as elevated respiration rate or panting, but continuous visual monitoring is labour intensive, time consuming and subjective. Therefore, different weather-based indices have been developed such as the temperature humidity index (THI) and heat load index (HLI) which are commonly used weather-based indices for monitoring cattle heat stress at commercial level. However, the thermal comfort level of cattle based on weather-based indices has limited use at a microclimatic and individual animal level. Varying sensor-based approaches have shown promise to shift the focus of heat stress management to the individual level. Monitoring individual animal response and mitigation strategies for isolated heat-susceptible cattle could save on heat management costs whilst improving animal welfare and productivity. Here we review the technologies that enable automatic, continuous, and real-time cattle heat stress monitoring and mitigation under commercial conditions. Future platforms for autonomous monitoring and mitigation of heat stress in cattle are likely to be based on minimally-invasive smart technologies either singly, or in an integrated system, enabling real-time solutions to animal responses under various production systems and environmental conditions.

show abstract

Using a CNN-LSTM for basic behaviors detection of a single dairy cow in a complex environment

Cited by 53 publications

References 36 publications

Data Augmentation for Inertial Sensor Data in CNNs for Cattle Behavior Classification

Data Augmentation for Inertial Sensor Data in CNNs for Cattle Behavior Classification

Intelligent Perception-Based Cattle Lameness Detection and Behaviour Recognition: A Review

Automated Monitoring of Cattle Heat Stress and Its Mitigation

Contact Info

Product

Resources

About