MOOnitor: An IoT based multi-sensory intelligent device for cattle activity monitoring
TL;DR: MOOnitor as discussed by the authors is a neck-mounted intelligent IoT device for cattle monitoring, which facilitates classification of salient activities of cattle through appropriately positioned sensors, including a temperature sensor, a GPS module, and a 3-axis accelerometer.
Abstract: Continuous activity monitoring of dairy cattle is essential to acquire a comprehensive knowledge on health and well-being of the animals. In this research, we have reported the development and deployment of "MOOnitor", a neck-mounted intelligent IoT device for cattle monitoring. The device facilitates classification of salient activities of cattle through appropriately positioned sensors. MOOnitor is an integration of a temperature sensor, a global positioning system (GPS) module, and a 3-axis accelerometer in a lightweight enclosure, which is attached to a halter that allows transmission of data to an IoT server using a microcontroller and a cellular GSM module. After acquiring the necessary sensory information, the most significant features were strategically extracted for enhanced data interpretation. Thereafter, optimally tuned eXtreme Gradient Boosting (XGBoost) and Random Forests classifiers were implemented to classify activities like ‘standing’, ‘lying’, ‘standing and ruminating’, ‘lying and ruminating’, ‘walking’, and ‘walking and grazing’. The performances of the two classifiers towards identification of different cattle activities were compared in terms of accuracy. Furthermore, the importance of using a temperature sensor and a GPS module in addition to an accelerometer in cattle activity recognition could be justified. An overall classification accuracy as high as ~97% was achieved using the XGBoost based classifier. In addition, accuracy, precision, sensitivity and specificity for standing (0.98, 0.97, 0.97, 0.98), lying (0.97, 0.90, 1, 0.96), standing and ruminating (0.99, 1, 0.97, 1), lying and ruminating (0.99, 1, 0.83, 1), walking (1, 1, 1, 1), and walking and grazing (0.99, 1, 0.75, 1) shows the suitability of the proposed method in effective cattle activity monitoring. Since cattle activity states are indicative of various factors such as estrous and several diseases like mastitis, foot-and-mouth disease, etc, the MOOnitor may be used for early detection of these conditions in addition to general health monitoring.
Citations
More filters
TL;DR: This paper investigates the usability of IoT and WSN for smallholder agriculture applications and identifies a significant future opportunity to design and implement affordable and reliable data acquisition tools and frameworks, with a possible integration of citizen science.
Abstract: Agriculture is the economy’s backbone for most developing countries. Most of these countries suffer from insufficient agricultural production. The availability of real-time, reliable and farm-specific information may significantly contribute to more sufficient and sustained production. Typically, such information is usually fragmented and often does fit one-on-one with the farm or farm plot. Automated, precise and affordable data collection and dissemination tools are vital to bring such information to these levels. The tools must address details of spatial and temporal variability. The Internet of Things (IoT) and wireless sensor networks (WSNs) are useful technology in this respect. This paper investigates the usability of IoT and WSN for smallholder agriculture applications. An in-depth qualitative and quantitative analysis of relevant work over the past decade was conducted. We explore the type and purpose of agricultural parameters, study and describe available resources, needed skills and technological requirements that allow sustained deployment of IoT and WSN technology. Our findings reveal significant gaps in utilization of the technology in the context of smallholder farm practices caused by social, economic, infrastructural and technological barriers. We also identify a significant future opportunity to design and implement affordable and reliable data acquisition tools and frameworks, with a possible integration of citizen science.
9 citations
TL;DR: A machine learning approach aiming at eliminating the influence of initial pressure on the identification of rumination and eating behaviors is proposed and will contribute to the standardized application and promotion of the noseband pressure sensors.
Abstract: Automatic monitoring of feeding behavior especially rumination and eating in cattle is important to keep track of animal health and growth condition and disease warnings. The noseband pressure sensor is not only able to accurately sense the pressure change of the cattle's jaw movements, which can directly reflect the cattle's chewing behavior, but also has strong resistance to interference. However, it is difficult to keep the same initial pressure while wearing the pressure sensor, and this will pose a challenge to process the feeding behavior data. This article proposed a machine learning approach aiming at eliminating the influence of initial pressure on the identification of rumination and eating behaviors. The method mainly used the local slope to obtain the local data variation and combined Fast Fourier Transform (FFT) to extract the frequency-domain features. Extreme Gradient Boosting Algorithm (XGB) was performed to classify the features of rumination and eating behaviors. Experimental results showed that the local slope in combination with frequency-domain features achieved an F1 score of 0.96, and recognition accuracy of 0.966 in both rumination and eating behaviors. Combined with the commonly used data processing algorithms and time-domain feature extraction method, the proposed approach improved the behavior recognition accuracy. This work will contribute to the standardized application and promotion of the noseband pressure sensors.
4 citations
TL;DR: In this paper , the authors highlight a key concern that occurs in the design and validation of IoT-based systems created for monitoring grazing cows in extensive agricultural systems, since they have many more, and more complicated, problems than indoor farms.
Abstract: Animal welfare is becoming an increasingly important requirement in the livestock sector to improve, and therefore raise, the quality and healthiness of food production. By monitoring the behaviour of the animals, such as feeding, rumination, walking, and lying, it is possible to understand their physical and psychological status. Precision Livestock Farming (PLF) tools offer a good solution to assist the farmer in managing the herd, overcoming the limits of human control, and to react early in the case of animal health issues. The purpose of this review is to highlight a key concern that occurs in the design and validation of IoT-based systems created for monitoring grazing cows in extensive agricultural systems, since they have many more, and more complicated, problems than indoor farms. In this context, the most common concerns are related to the battery life of the devices, the sampling frequency to be used for data collection, the need for adequate service connection coverage and transmission range, the computational site, and the performance of the algorithm embedded in IoT-systems in terms of computational cost.
2 citations
TL;DR: This article combines the wavelets function with the planning algorithm and proposes a dynamic programming algorithm, which removes the redundant monitoring data in turn and clusters the distortion monitoring data with the wavelet function, which improves the accuracy and computational efficiency of the algorithm and gives full play to the monitoring of intelligence.
Abstract: With the continuous development of the national economy and scientific productivity, urban construction and people's living standards are also getting higher and higher. Although people enjoy increasingly convenient life, the demand for intelligence is getting higher and higher. Digital intelligent equipment has the functions of data collection, calculation and analysis, diagnostic and early warning, and communication functions. Analyze the status quo and existing problems of the development of intelligent equipment, as well as analyze and research key monitoring technologies in the use and development of digital intelligent equipment and provide optimal solutions for intelligent equipment hardware development requirements, software development, and model algorithms. Intelligent equipment monitoring is related to all aspects of people's livelihood, and its intelligent development is related to the public role in this field in the future. Accurate results of monitoring can provide data support for schools, research institutions, the public, and the government. At the same time, it is also an important basis for formulating social policies. At present, the commonly used monitoring method usually adopts time series algorithm. Through literature review, it is found that the algorithm has the problem of distortion of correct data, which affects the accuracy of monitoring results. Based on the above reasons, this article combines the wavelet function with the planning algorithm and proposes a dynamic programming algorithm, which removes the redundant monitoring data in turn and clusters the distortion monitoring data with the wavelet function, which improves the accuracy and computational efficiency of the algorithm and gives full play to the monitoring of intelligence. The simulation results using MATLAB show that the planning algorithm can eliminate 90% of redundant monitoring data and improve the extraction rate of characteristic monitoring data. At the same time, the accuracy of the planning algorithm reaches 95%, and the calculation time is less than 25 s, which is better than the static planning algorithm. Therefore, the dynamic programming algorithm can better utilize the intelligence, convenience, and efficiency of the equipment to optimize the monitoring model.
TL;DR: In this article , a fixed-length feature extraction (FLFE) algorithm was used to analyze the respiratory rate and other factors of a ruminant while eating, which yielded a prediction accuracy of 94.5% for respiratory rate of the cow when compared to healthy animals.
Abstract: The presence of many animals with different body types and characteristics necessitates the need for cattle husbandry system. Especially in cows, elevated heart rates have been linked to symptoms of stress, such as sweating and anguish, and they can provide valuable information for monitoring systems. There are many ways to measure a cow’s pulse rate. In the current study, cow’s pulmonary function, cardiovascular system, contemplation rates and durations have been measured using biomedical sensors. An electro cardiogram-based sensor (ECS) approach for noninvasively monitoring the ruminant’s ingestive activity has also been developed in this study. A sensor adapter is used to sample the chewing surface Electrocardiogram (ECG) signal from ruminant animals’ masseter muscles while eating. When it comes to this, an intra-ruminal real-time sensor is designed to get accurate information on the ruminal activity of cows while grazing. A Fixed-Length Feature Extraction (FLFE) algorithm is used to analyze the respiratory rate and other factors. Four segmentation methods that have been tested and used to split the chewing signal automatically are Blind Fragmentation (BF), Fixed Duration Peak-Centered Segmentation (FDPCS), Double Onset Segmentation (DOS) and Fixed-Length Feature Extraction (FLFE) algorithm. Digital components have been integrated into an IoT-enabled digital platform for commercial use. The obtained ECG signal is extracted and segmented to analyze the respiratory rate and split the signal chewing state. Its main goal is to automate some monotonous animal care activities using Internet of Things and Artificial Intelligence (AI) to ensure better care and management of animals. The FLFE algorithm yielded a prediction accuracy of 94.5% for the respiratory rate of the cow when compared to healthy animals.
References
More filters
[...]
TL;DR: Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the forest, and are also applicable to regression.
Abstract: Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, aaa, 148–156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.
79,257 citations
Journal Article•
TL;DR: This paper shows empirically and theoretically that randomly chosen trials are more efficient for hyper-parameter optimization than trials on a grid, and shows that random search is a natural baseline against which to judge progress in the development of adaptive (sequential) hyper- parameter optimization algorithms.
Abstract: Grid search and manual search are the most widely used strategies for hyper-parameter optimization. This paper shows empirically and theoretically that randomly chosen trials are more efficient for hyper-parameter optimization than trials on a grid. Empirical evidence comes from a comparison with a large previous study that used grid search and manual search to configure neural networks and deep belief networks. Compared with neural networks configured by a pure grid search, we find that random search over the same domain is able to find models that are as good or better within a small fraction of the computation time. Granting random search the same computational budget, random search finds better models by effectively searching a larger, less promising configuration space. Compared with deep belief networks configured by a thoughtful combination of manual search and grid search, purely random search over the same 32-dimensional configuration space found statistically equal performance on four of seven data sets, and superior performance on one of seven. A Gaussian process analysis of the function from hyper-parameters to validation set performance reveals that for most data sets only a few of the hyper-parameters really matter, but that different hyper-parameters are important on different data sets. This phenomenon makes grid search a poor choice for configuring algorithms for new data sets. Our analysis casts some light on why recent "High Throughput" methods achieve surprising success--they appear to search through a large number of hyper-parameters because most hyper-parameters do not matter much. We anticipate that growing interest in large hierarchical models will place an increasing burden on techniques for hyper-parameter optimization; this work shows that random search is a natural baseline against which to judge progress in the development of adaptive (sequential) hyper-parameter optimization algorithms.
6,935 citations
TL;DR: Based on this study, the best variable selection methods for most datasets are Jiang's method and the method implemented in the VSURF R package, and for datasets with many predictors, the methods implement in the R packages varSelRF and Boruta are preferable due to computational efficiency.
Abstract: Random forest classification is a popular machine learning method for developing prediction models in many research settings. Often in prediction modeling, a goal is to reduce the number of variables needed to obtain a prediction in order to reduce the burden of data collection and improve efficiency. Several variable selection methods exist for the setting of random forest classification; however, there is a paucity of literature to guide users as to which method may be preferable for different types of datasets. Using 311 classification datasets freely available online, we evaluate the prediction error rates, number of variables, computation times and area under the receiver operating curve for many random forest variable selection methods. We compare random forest variable selection methods for different types of datasets (datasets with binary outcomes, datasets with many predictors, and datasets with imbalanced outcomes) and for different types of methods (standard random forest versus conditional random forest methods and test based versus performance based methods). Based on our study, the best variable selection methods for most datasets are Jiang's method and the method implemented in the VSURF R package. For datasets with many predictors, the methods implemented in the R packages varSelRF and Boruta are preferable due to computational efficiency. A significant contribution of this study is the ability to assess different variable selection techniques in the setting of random forest classification in order to identify preferable methods based on applications in expert and intelligent systems.
446 citations
TL;DR: To achieve this, it is necessary that production systems become market-orientated, better regulated in cases, and socially acceptable so that the right mix of incentives exists for the systems to intensify.
Abstract: Livestock play a significant role in rural livelihoods and the economies of developing countries. They are providers of income and employment for producers and others working in, sometimes complex, value chains. They are a crucial asset and safety net for the poor, especially for women and pastoralist groups, and they provide an important source of nourishment for billions of rural and urban households. These socio-economic roles and others are increasing in importance as the sector grows because of increasing human populations, incomes and urbanisation rates. To provide these benefits, the sector uses a significant amount of land, water, biomass and other resources and emits a considerable quantity of greenhouse gases. There is concern on how to manage the sector's growth, so that these benefits can be attained at a lower environmental cost. Livestock and environment interactions in developing countries can be both positive and negative. On the one hand, manures from ruminant systems can be a valuable source of nutrients for smallholder crops, whereas in more industrial systems, or where there are large concentrations of animals, they can pollute water sources. On the other hand, ruminant systems in developing countries can be considered relatively resource-use inefficient. Because of the high yield gaps in most of these production systems, increasing the efficiency of the livestock sector through sustainable intensification practices presents a real opportunity where research and development can contribute to provide more sustainable solutions. In order to achieve this, it is necessary that production systems become market-orientated, better regulated in cases, and socially acceptable so that the right mix of incentives exists for the systems to intensify. Managing the required intensification and the shifts to new value chains is also essential to avoid a potential increase in zoonotic, food-borne and other diseases. New diversification options and improved safety nets will also be essential when intensification is not the primary avenue for developing the livestock sector. These processes will need to be supported by agile and effective public and private institutions.
351 citations
TL;DR: An efficient machine learning method, random forests in combination with extreme gradient boosting (XGBoost), is used to establish the data-driven wind turbine fault detection framework that is robust to various wind turbine models including offshore ones in different working conditions.
Abstract: Wind energy has seen great development during the past decade. However, wind turbine availability and reliability, especially for offshore sites, still need to be improved, which strongly affect the cost of wind energy. Wind turbine operational cost is closely depending on component failure and repair rate, while fault detection and isolation will be very helpful to improve the availability and reliability factors. In this paper, an efficient machine learning method, random forests (RFs) in combination with extreme gradient boosting (XGBoost), is used to establish the data-driven wind turbine fault detection framework. In the proposed design, RF is used to rank the features by importance, which are either direct sensor signals or constructed variables from prior knowledge. Then, based on the top-ranking features, XGBoost trains the ensemble classifier for each specific fault. In order to verify the effectiveness of the proposed approach, numerical simulations using the state-of-the-art wind turbine simulator FAST are conducted for three different types of wind turbines in both the below and above rated conditions. It is shown that the proposed approach is robust to various wind turbine models including offshore ones in different working conditions. Besides, the proposed ensemble classifier is able to protect against overfitting, and it achieves better wind turbine fault detection results than the support vector machine method when dealing with multidimensional data.
341 citations
Related Papers (1)
14 Dec 2022