Papers - THI THI ZIN
-
Research on Individual Identification of Walking Cows Using a 3D Camera Reviewed International journal
Y. Shiihara, Thi Thi Zin, M. Aikawa, I. Kobayashi
Lecture Notes in Electrical Engineering 1322 LNEE 73 - 83 2025.2
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：Lecture Notes in Electrical Engineering
This research focuses on identifying individual cows based on features from their back regions captured by 3D cameras, aiming to enhance management efficiency and reduce labor burdens. The methodology involves using 3D cameras to collect data on cows walking through a milking parlor. The captured data is processed to extract specific features such as pixel counts and volume from the cow’s back region. Various classification methods, including SVM, k-NN, decision trees, and random forests, are employed to identify individual cows. Experimental results demonstrated that the random forest classifier achieved the highest accuracy at 95%, outperforming other methods. The study highlights the limitations of current RFID-based systems, such as cost and stress on animals, and presents a non-contact alternative that reduces labor and improves accuracy.
-
Enhancing Fetal Monitoring through Digital Twin Technology and Entropy-Based Fetal Heart Rate Variability Analysis Reviewed International journal
Tunn Cho Lwin, Thi Thi Zin, Pyae Phyo Kyaw, Pyke Tin, E. Kino and T. Ikenoue
International Journal of Innovative Computing, Information and Control 21 ( 1 ) 185 - 196 2025.2
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：International Journal of Innovative Computing, Information and Control
In fetal healthcare, Digital Twin Technology (DTT) offers a powerful tool for simulating fetal physiological conditions, enabling continuous, real-time monitoring and predictive analysis. This study explores the integration of DTT with entropy-based analysis of fetal heart rate variability (FHRV) to enhance fetal monitoring. Utilizing a dataset of 585 fetal electrocardiogram (ECG) recordings collected via scalp electrode monitoring during delivery, we computed entropy measures such as Markov entropy and multiscale entropy to assess fetal status. The results demonstrate that these entropy measures provide significant information regarding fetal well-being status. Moreover, the calculated entropy values correlate strongly with umbilical cord blood gas parameters. This correlation suggests that entropy-based FHRV analysis, combined with DTT, can serve as an effective and reliable method for improving the accuracy of fetal health monitoring and predicting fetal well-being as delivery approaches.
-
Automated Cattle Monitoring System for Calving Time Prediction Using Trajectory Data Embedded Time Series Analysis Reviewed International journal
Wai Hnin Eaindrar Mg, Thi Thi Zin, Pyke Tin, M. Aikawa, K. Honkawa, Y. Horii
IEEE Open Journal of the Industrial Electronics Society 6 1 - 19 2025.1
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：IEEE Open Journal of the Industrial Electronics Society
This research introduces an automated system for cattle monitoring and calving time prediction, utilizing trajectory data embedded with time-series analysis. Designed for large-scale farms, our system offers continuous 12-hour monitoring, ensuring precise capture of cattle movements. By utilizing time series analysis on the trajectory data, our system predicts calving events in advance, effectively distinguishing between abnormal (requiring human assistance) and normal (not requiring assistance) for each cow. We utilized 360-degree surveillance cameras to provide comprehensive coverage without disturbing the cattle's natural behavior. We employed tailored versions of the Detectron2 and YOLOv8 models to achieve efficient and precise cattle detection, comparing their performance in terms of missed detections and false detections. For tracking, we used our Customized Tracking Algorithm (CTA), which minimizes ID switching and ensures continuous identification even in challenging conditions such as occlusions. While some ID switching errors still occur over extended tracking periods, we integrated tracking and identification to further optimize the handling of track IDs and global IDs. Our system incorporates a 4-hour forecasting of cattle movement using Euclidean Fluctuating Summation (EFS) feature combined with our custom Long Short-Term Memory (LSTM) model. Experimental results demonstrate a detection accuracy of 98.70%, tracking and identification accuracy of 99.18%, and forecasting with an average error rate of 14.07%. Furthermore, the system accurately classifies cattle as either normal or abnormal and predicts calving events a 4-hour in advance using the EFS feature, comparing its performance with various machine learning algorithms. The system's seamless integration significantly enhances farm management and animal welfare.
-
Automated system for calving time prediction and cattle classification utilizing trajectory data and movement features. Reviewed International journal
Wai Hnin Eaindrar Mg, Thi Thi Zin, Pyke Tin, M. Aikawa, K. Honkawa, Y. Horii
Scientific reports 15 ( 1 ) 2378 2025.1
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：Scientific reports
Accurately predicting the calving time in cattle is essential for optimizing livestock management and ensuring animal welfare. Our research focuses on developing a robust system for calving cattle classification and calving time prediction, utilizing 12-h trajectory data for 20 cattle. Our system classifies cattle as abnormal (requiring human assistance) or normal (not requiring assistance) and predicts calving times based on their individual behaviors. We employed a tailored YOLOv8 model for efficient and precise cattle detection, effectively filtering out noise such as people and trucks. Our Customized Tracking Algorithm (CTA) maintains continuous identity tracking for each cow, enabling accurate re-identification even during occlusions. To minimize some ID switching errors over extended tracking periods, we integrated IDs optimization in the CTA utilizing Global IDs identification. We extracted and compared three total movement features for classifying cattle as abnormal or normal. For predicting calving times for each cow, we utilized and compared three cumulative movement features. Our system is fully automated, detecting and tracking all 20 cattle continuously for 12 h without manual assistance, and achieving an overall accuracy of 99%. By comparing three features derived from the trajectory tracking data for each point in a frame, we achieve 100%, 95%, 85% accuracy in classifying cattle as abnormal or normal and predict their calving times with a precision of within the next 6 h, within the next 9 h, within the next 8 h, respectively. Our system enables farmers to provide timely assistance, ensuring the health and safety of both the cow and the calf. Furthermore, it aids in optimizing resource allocation and enhancing overall farm efficiency, emphasizing the critical importance of calving time prediction in sustainable livestock farming.
-
A study on Depth Camera-Based Estimation of Elderly Patient Actions Reviewed
NAKASHIMA Remon, Thi Thi Zin, KONDO Kazuhiro, Watanabe Shinji
Proceedings of the Annual Conference of Biomedical Fuzzy Systems Association 37 ( 0 ) 46 - 52 2024.12
Authorship：Corresponding author Language：Japanese Publishing type：Research paper (conference, symposium, etc.) Publisher：Biomedical Fuzzy Systems Association
-
AI-enhanced real-time cattle identification system through tracking across various environments Reviewed International journal
Su Larb Mon, T. Onizuka, Pyke Tin, M. Aikawa, I. Kobayashi, Thi Thi Zin
Scientific Reports 14 ( 1 ) 17779 2024.12
Authorship：Lead author, Last author Language：English Publishing type：Research paper (scientific journal) Publisher：Scientific Reports
Video-based monitoring is essential nowadays in cattle farm management systems for automated evaluation of cow health, encompassing body condition scores, lameness detection, calving events, and other factors. In order to efficiently monitor the well-being of each individual animal, it is vital to automatically identify them in real time. Although there are various techniques available for cattle identification, a significant number of them depend on radio frequency or visible ear tags, which are prone to being lost or damaged. This can result in financial difficulties for farmers. Therefore, this paper presents a novel method for tracking and identifying the cattle with an RGB image-based camera. As a first step, to detect the cattle in the video, we employ the YOLOv8 (You Only Look Once) model. The sample data contains the raw video that was recorded with the cameras that were installed at above from the designated lane used by cattle after the milk production process and above from the rotating milking parlor. As a second step, the detected cattle are continuously tracked and assigned unique local IDs. The tracked images of each individual cattle are then stored in individual folders according to their respective IDs, facilitating the identification process. The images of each folder will be the features which are extracted using a feature extractor called VGG (Visual Geometry Group). After feature extraction task, as a final step, the SVM (Support Vector Machine) identifier for cattle identification will be used to get the identified ID of the cattle. The final ID of a cattle is determined based on the maximum identified output ID from the tracked images of that particular animal. The outcomes of this paper will act as proof of the concept for the use of combining VGG features with SVM is an effective and promising approach for an automatic cattle identification system
-
Predictive modeling of cattle calving time emphasizing abnormal and normal cases by using posture analysis Reviewed International journal
May Phyu Khin, Pyke Tin, Y. Horii, Thi Thi Zin
Scientific Reports 14 ( 1 ) 31871 2024.12
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：Scientific Reports
Accurate calving time prediction plays a critical role in ensuring the well-being of both mother and calf during parturition. Challenges during the calving process, particularly in abnormal cases, often necessitate human intervention to prevent potentially fatal outcomes. This study proposes a novel system for automated prediction of normal and abnormal cattle calving cases based on posture analysis. By analyzing changes in posture and identifying specific posture types exhibited by cattle, the system aims to provide early warnings of impending calving events, enabling timely intervention and risk mitigation measures. Leveraging advanced computer vision techniques, particularly the Mask R-CNN from the Detectron2 detection and the YOLOv8-pose classification method known for their efficient training time and overall accuracy, the system analyzes the frequency of posture changes and key postures like sitting, standing, feeding, sitting with extended legs, and tail-raised to predict calving cases with high precision. We discovered that the “sitting with leg extended” posture is a crucial indicator for abnormal calving events. By incorporating this posture into the classification process, the system aims to achieve high accuracy in predicting both normal and abnormal calving timeframes. Additionally, the system differentiates between normal and abnormal calving patterns by analyzing posture sequences leading up to parturition, focusing on timeframes such as 30 min, 1 h, and 2 h pre-calving. This comprehensive analysis aids in identifying potential calving complications and enables the implementation of proactive management strategies. By offering insights into optimal methods for predicting specific postures and optimizing calving time management practices, this research contributes to the field of precision livestock farming, ultimately enhancing animal welfare and reducing calving-related risks.
-
ENHANCING FACTORY EFFICIENCY: DATA-DRIVEN ANALYSIS WITH WORKER DETECTION AND TRACKING SYSTEM Reviewed International journal
I. Hidaka, S. Inoue, T. Ishikawa, H. Tamura, Thi Thi Zin
ICIC Express Letters 18 ( 12 ) 1327 - 1337 2024.12
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
As Japan’s population continues to decline, one challenge for small and medium-sized enterprises (SMEs) is the decline in productivity due to a shortage of employees. As a result, many small- and medium-scale factories are introducing AI and IoT to automate their operations so that they can handle them with fewer employees. However, some factories are not automating because it is more economical for employees to do the work directly. Such companies can produce more economic benefits than they do now with less labor by eliminating waste from their current work and improving work efficiency. Therefore, in this paper, we propose a system to detect and track workers in a factory using a 4K camera to obtain trajectories (lines of movement) of each work group and acquire data to be used for improving work efficiency. To determine the work group, markers are attached to the top of workers’ helmets. The proposed system is simulated with a dataset taken by us in a real factory, and the average accuracy of helmet detection and tracking is 92.9% and 86.0%, respectively. The proposed system allows us to visually see the trajectory of workers, which will lead to decision making on staffing and work process changes and improve work efficiency.
-
Unobtrusive Elderly Action Recognition with Transitions Using CNN-RNN Reviewed
Ye Htet, Thi Thi Zin, H. Tamura, K. Kondo, E. Chosa
Journal of Signal Processing 28 ( 6 ) 315 - 319 2024.11
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：信号処理学会
This study addresses the efficient recognition of elderly people's daily actions, emphasizing transition states, using privacy-preserving depth data and deep learning algorithms. Stereo-depth cameras collect data from an elder care center, ensuring privacy by capturing only depth information without revealing identifiable details. The research investigates spatial and temporal features in movement patterns by employing a Convolutional Neural Network (CNN) for transfer learning on segmented person image sequences to extract spatial features, while a Recurrent Neural Network (RNN) decoder extracts temporal features. The proposed study evaluated various CNN and RNN integrated architectures, assessing algorithmic performance on real-world data from three elderly participants. Experimental outcomes reveal the best model achieving 95% overall accuracy for all actions and an average accuracy of over 80% for classifying transition states. Beyond accuracy, comprehensive evaluation includes precision, recall, and F1-score, offering a thorough assessment of the developed algorithm's practical effectiveness on real-world data.
DOI： 10.2299/jsp.28.315
-
A STUDY OF THE EARLY DETECTION OF OTITIS MEDIA IN CALVES WITH TWO TYPES OF CAMERAS Reviewed International journal
T. Nishiyama, K. Shiiya, M. Aikawa, I. Kobayashi, Thi Thi Zin
ICIC Express Letters, Part B: Applications 15 ( 11 ) 1183 - 1191 2024.11
Authorship：Last author Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
Calves tend to be more susceptible to infections than adult cattle. This may be due to a less mature immune system, stress from the rearing environment, and other factors. Early detection of disease can help prevent deterioration and the spread of infection. Therefore, in this study, we proposed to investigate the early detection of mycoplasma otitis media in a non-contact manner using RGB and thermal imaging cameras. We then conducted experiments at the Sumiyoshi Field of the University of Miyazaki to confirm the effectiveness of the proposed method.
-
Analyzing Parameter Patterns in YOLOv5-based Elderly Person Detection Across Variations of Data Reviewed International journal
Ye Htet, Thi Thi Zin, Pyke Tin, H. Tamura, K. Kondo, S. Watanabe, E. Chosa
Proceedings - 2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems, MASS 2024 629 - 634 2024.10
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：Proceedings - 2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems, MASS 2024
This study investigates the impact of data variations on parameter patterns within a YOLOv5-based elderly person detection model. We explore how changes in camera settings and environmental factors influence the model's parameters. Our research questions focus on how these variations affect parameter patterns, model robustness, and generalizability. We aim to identify the parameters and layers of the model most susceptible to variations and develop strategies to improve the model's performance across various datasets. The experiment involves collecting data from eight elderly participants in real-life elder care facility settings. During data acquisition, only depth images are recorded using stereo-depth cameras to protect privacy. After that, we train individual YOLOv5 models for each dataset through hyperparameter tuning and transfer learning. Optimal hyperparameters and the sensitive convolutional layer of each model are then compared. Class Activation Maps (CAMs) are utilized to visualize the network's focus, followed by analysis of weight distributions and correlation to identify parameter patterns. The findings will provide valuable insights for improving elderly person detection models for smart health care and their robustness to real-world variations.
-
Fusion of Strategic Queueing Theory and AI for Smart City Telecommunication System Reviewed International journal
Thi Thi Zin, Aung Si Thu Moe, Cho Nilar Phyo, Pyke Tin
Proceedings - 2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems, MASS 2024 653 - 657 2024.10
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：Proceedings - 2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems, MASS 2024
This paper explores the innovative intersection of queueing theory and artificial intelligence (AI), addressing emerging challenges for Smart City Telecommunication System. We propose a novel approach that integrates queueing theory with AI methodologies, particularly artificial neural networks (ANNs), to optimize strategic parameters in Markov Decision Process (MDP) problems in communication systems. These parameters include server numbers, customer waiting times, queue lengths, and other critical metrics. Our fusion approach demonstrates high efficacy and indicates significant advancements in applying machine learning to complex queueing theory issues. The study underscores the practical applications of this integration across various domains. We provide real-world examples illustrating the strategic use of AI-enhanced queueing models in improving user experiences and optimizing system efficiencies. Our discussions cover the benefits of intelligent resource allocation, dynamic load balancing, and adaptive service time optimization. We extend our exploration through specific case studies demonstrating the efficacy of this integration in industries such as telecommunications, healthcare, and transportation. For instance, in telecommunications, AI-driven queue management systems can dynamically allocate bandwidth to ensure optimal service quality. In healthcare, queueing models enhanced by AI can improve patient flow and reduce wait times, potentially leading to better healthcare outcomes. In this paper, we investigate the application of deep learning models to queueing theory, specifically focusing on [mention specific problem or application, e.g., 'forecasting queue lengths in real-time systems']. We implement and evaluate several deep learning architectures to determine their effectiveness in modeling and predicting queue dynamics. The paper concludes by emphasizing the necessity for continued interdisciplinary research, encouraging collaboration between AI experts and queueing theorists. This synergy is essential for unlocking new potential and addressing the increasingly complex challenges posed by modern systems. By fostering such collaboration, we anticipate significant breakthroughs that will transform both fields, leading to more efficient, adaptive, and intelligent systems capable of meeting future demands across various sectors.
-
AUTOMATED CATTLE DETECTION USING MASK R-CNN AND IOU-BASED TRACKING WITH A SINGLE SIDE-VIEW CAMERA Reviewed International journal
Bo Bo Myint, T. Onizuka, Pyke Tin, M. Aikawa, I. Kobayashi, Thi Thi Zin
International Journal of Innovative Computing, Information and Control 20 ( 5 ) 1439 - 1447 2024.10
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：International Journal of Innovative Computing, Information and Control
In precision livestock farming, the early detection of lameness in cattle is an extremely important aspect of effective breeding management. Timely identification of lameness not only facilitates prompt and cost-efficient treatment but also plays a crucial role in avoiding possible future diseases. This study emphasizes the significance of intelligent visual perception systems for lameness detection in dairy cattle, particularly in the lane between from Milking Parlor to Cattle Barn. To address the cattle lameness issue, we employ an advanced deep learning, and image processing technique, i.e., Mask R-CNN from Detectron2 to detect and identify cattle regions for feature extraction of lameness detection. On the other hand, cattle tracking using IoU is also an important part of data accumulation for lameness classification. The results of this study contribute to ongoing efforts in precision animal husbandry and demonstrate the potential of intelligent visual recognition systems for early lameness detection.
-
A 3D CAMERA APPROACH TO EVALUATING BODY CONDITION SCORE IN WALKING DAIRY COWS Reviewed International journal
M. Chikunami, Thi Thi Zin, M. Aikawa, I. Kobayashi
ICIC Express Letters, Part B: Applications 15 ( 10 ) 1089 - 1097 2024.10
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
Body Condition Score (BCS) is an important index for assessing body fat accumulation in cattle and plays a crucial role in managing cattle productivity, feeding efficiency, and overall health. Currently, BCS evaluations predominantly rely on visual assessment and palpation by specialized personnel, which is time-consuming and laborintensive. Consequently, many farms refrain from utilizing BCS for cattle management. Previous studies have focused on BCS evaluation of stationary dairy cows in rotary parlors, but this approach is not feasible for small and medium-sized livestock producers lacking such facilities. To enable BCS management for dairy cows on any farm, we propose a system utilizing image processing technology for evaluating cows while walking. In this system, 3D cameras are employed to capture images, and an evaluation model is constructed using feature extraction and multiple regression analysis. This model allowed the evaluation of cows with large BCS within an error margin of 0.25.
-
A Markovian Game Theoretic Framework for Analysing a Queueing System with Multiple Servers Reviewed International journal
Pyke Tin, Thi Thi Zin
2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024 2024.7
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024
This paper introduces a Markov game theoretic framework designed to analyze a Markovian queueing system equipped with multiple servers. Our focus lies in modelling a message transmission system, where messages traverse various transmission options, each associated with a cost and governed by a decision process. The primary objective is to investigate the impact of cooperation and communication, or their absence, among servers. The inherent uncertainty regarding the characteristics of the available transmission alternatives is mathematically captured through a Markovian game formulation. Within this framework, we quantify the inefficiency resulting from the self-interested management of individual servers and the associated loss attributed to the decision-making process. Our analysis encompasses diverse scenarios of signaling exchange among servers, providing valuable insights into the system's behavior under varying conditions.
-
Enhancing Fetal Heart Rate Monitoring Through Digital Twin Technology Reviewed International journal
Tunn Cho Lwin, Thi Thi Zin, Pyke Tin, T. Ikenoue, E. Kino
2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024 2024.7
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024
In the era of digital twin technology (DTT), which involves creating virtual replicas of physical systems, healthcare applications have seen a surge in innovation. Fetal heart rate monitoring, a rapidly advancing field within healthcare, is benefiting greatly from the effective implementation of DTTs. Digital twin, supported by Artificial Intelligence (AI), Virtual Reality (VR), and Extended Virtuality (EV), have already demonstrated significant impact in entertainment, gaming, and media sectors. This paper explores and analyzes fetal heart rate monitoring systems during labor using digital twin technologies. During labor, there is an elevated risk of developing non-communicable diseases (NCDs) such as diabetes, cardiovascular diseases, cancer, and chronic respiratory diseases. The developmental origins of health and disease hypothesis posits that environmental conditions during fetal and early postnatal development have enduring effects on growth, structure, and metabolism, ultimately influencing long-term health and well-being. Understanding and effectively monitoring fetal heart rate dynamics during labor using digital twin technologies can provide valuable insights into maternal and fetal health. The findings underscore the value of entropy analysis in FHRV assessment, presenting a pioneering predictor of fetal health via ECG data. Moreover, it reveals the current landscape of fetal heart rate monitoring, discusses the integration of digital twins, and proposes future directions for optimizing healthcare outcomes in the digital era.
-
A Stochastics Branching Process Model for Analyzing Rumor Spreading in Social Media Networks Reviewed International journal
Thi Thi Zin, Pyke Tin, H. Hama
2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024 2024.7
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2024 IEEE Gaming, Entertainment, and Media Conference, GEM 2024
This paper introduces a stochastics branching process model as an analytical tool to study the dynamics of rumor spreading in social media networks. Employing a specialized first-order conditional probability generating function method, we derive a second-order linear regressive equation to characterize the intricate dynamics of rumor propagation. The validation of our model's accuracy is established through a comparative analysis between numerical computations and Monte Carlo simulations. Additionally, we present a derived threshold condition for the spread control factor, also known as the reproduction rate. Numerical simulation results demonstrate that the interactions and responses within social platforms contribute to the rapid onset of rumors while simultaneously diminishing the maximum density of spreaders and the overall scale of the rumors. To provide further insights, we explore analogies between rumor-spreading models and epidemic models. In conclusion, our proposed stochastics branching process model not only enhances our understanding of rumor spreading in social media networks but also offers a valuable framework for investigating the interplay between various factors influencing the dynamics of information diffusion.
-
Precision Livestock Tracking: Advancements in Black Cattle Monitoring for Sustainable Agriculture Reviewed
Su Myat Noe, Thi Thi Zin, Pyke Tin, I. Kobayashi
Journal of Signal Processing 28 ( 4 ) 179 - 182 2024.7
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：信号処理学会
Utilizing computer vision for animal behavior monitoring provides significant benefits by minimizing direct handling and capturing diverse traits through a single sensor. However, accurately identifying animals remains a challenge. To address this, this study introduces an innovative approach to monitor black cattle in dynamic agricultural environments to ensure their health welfare. By integrating advanced techniques like DETIC for automated labeling and YOLOv8 for real-time detection, the research emphasizes improving accuracy and robustness in tracking black cattle tracking within complex open ranch environments. Moreover, the customized ByteTrack model tailored for ranch scenarios significantly enhances cattle tracking across intricate landscapes. Achieving a mean Average Precision (mAP) of 0.901 and a Multi-Object Tracking Accuracy (MOTA) of average accuracy 92.185% of four videos, this approach appears to offer a viable resolution for conducting individual cattle behavior analysis experiments through the application of computer vision.
DOI： 10.2299/jsp.28.179
-
Identifying Black Cow Actions Using Kalman Filter Velocity and Multi-Stage Classification Reviewed
Cho Cho Aye, Thi Thi Zin, M. Aikawa, I. Kobayashi
Journal of Signal Processing 28 ( 4 ) 183 - 186 2024.7
Authorship：Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：信号処理学会
This study proposes an advanced camera-based monitoring system for individual black cows in closed farms. By leveraging computer vision and deep learning, the system identifies five key cow actions: eating, drinking, sitting, standing, and walking. A multi-stage approach classifies actions first as static (eating, drinking, sitting, standing) or dynamic (walking) categories based on Kalman Filter velocity information. Further classification distinguishes among four static actions. A Convolutional Neural Network (CNN) refines especially for sitting and standing. On the other hand, cow head regions and specific zone locations help distinguish eating and drinking. The system achieves an overall accuracy of 80% in long data sequences, demonstrating its potential for precision livestock farming.
DOI： 10.2299/jsp.28.183
-
Development of a real-time cattle lameness detection system using a single side-view camera Reviewed International journal
Bo Bo Myint, T. Onizuka, Pyke Tin, M. Aikawa, I. Kobayashi, Thi Thi Zin
Scientific reports 14 ( 1 ) 13734 2024.6
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：Scientific Reports
Recent advancements in machine learning and deep learning have revolutionized various computer vision applications, including object detection, tracking, and classification. This research investigates the application of deep learning for cattle lameness detection in dairy farming. Our study employs image processing techniques and deep learning methods for cattle detection, tracking, and lameness classification. We utilize two powerful object detection algorithms: Mask-RCNN from Detectron2 and the popular YOLOv8. Their performance is compared to identify the most effective approach for this application. Bounding boxes are drawn around detected cattle to assign unique local IDs, enabling individual tracking and isolation throughout the video sequence. Additionally, mask regions generated by the chosen detection algorithm provide valuable data for feature extraction, which is crucial for subsequent lameness classification. The extracted cattle mask region values serve as the basis for feature extraction, capturing relevant information indicative of lameness. These features, combined with the local IDs assigned during tracking, are used to compute a lameness score for each cattle. We explore the efficacy of various established machine learning algorithms, such as Support Vector Machines (SVM), AdaBoost and so on, in analyzing the extracted lameness features. Evaluation of the proposed system was conducted across three key domains: detection, tracking, and lameness classification. Notably, the detection module employing Detectron2 achieved an impressive accuracy of 98.98%. Similarly, the tracking module attained a high accuracy of 99.50%. In lameness classification, AdaBoost emerged as the most effective algorithm, yielding the highest overall average accuracy (77.9%). Other established machine learning algorithms, including Decision Trees (DT), Support Vector Machines (SVM), and Random Forests, also demonstrated promising performance (DT: 75.32%, SVM: 75.20%, Random Forest: 74.9%). The presented approach demonstrates the successful implementation for cattle lameness detection. The proposed system has the potential to revolutionize dairy farm management by enabling early lameness detection and facilitating effective monitoring of cattle health. Our findings contribute valuable insights into the application of advanced computer vision methods for livestock health management.