Papers - THI THI ZIN
-
Predicting calving time of dairy cows by exponential smoothing models Reviewed International coauthorship
Tunn Cho Lwin, Thi Thi Zin, Pyke Tin
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 322 - 323 2020.10
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
In dairy farming, calving time prediction is crucial because the calf loss rate during calving time is rising for many reasons. In this paper, we propose a time series model with exponential smoothing technique to predict the time of calving event occurs. By investigating the changes in behavior patterns a few days before the calving events, the proposed method can predict accurately the time of occurrence of calving events. To be specific, we model the number of changes in behaviors of standing and lying as a time series in hourly basis. We then employ the exponential smoothing techniques and survival probability with exponential distribution to make the prediction process. To confirm the proposed method, some experimental works are performed by using video records of 25 cows in calving pen from a large farm at Oita prefecture, Japan.
-
Image Processing and Statistical Analysis Approach to Predict Calving Time in Dairy Cows Reviewed
Swe Zar Maw, Thi Thi Zin, Pyke Tin
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 318 - 319 2020.10
Authorship：Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
An accurate prediction of calving time in dairy cows is one of the most important factors to make an optimal reproduction process in dairy farming. This paper proposes an image processing and statistical analysis approach to predict calving time in dairy cows. Specifically, we extract the behavior changes patterns of the expected cows by using simple effective motion history images (MHI) a few days before the occurrence of calving event from the video sequences taken in the maternity bans. We then classify extracted features with support vector machine (SVM) and analyze the behavior changes by using statistical method, Hidden Markov model (HMM) for prediction process. To confirm the validity of proposed method, we perform some experiments by installing 360-degree view cameras at the top of calving bans. At the first stage, we analyzed the behaviors of 25 dairy cows for 72 hours before giving birth. As a result, we find that the proposed method is promising.
-
Feature Detection and Classification of Cow Motion for Predicting Calving time Reviewed
Thi Thi Zin, Saw Zay Maung Maung, Pyke Tin, Y. Horii
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 305 - 306 2020.10
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
The monitoring and automatic detecting of cow behaviors is a key factor for predicting cow calving times. This paper describes the analysis of cow motion patterns by using 360 camera in order to identify various views of cow states. Firstly, Principle Component Analysis (PCA) is applied to solve the rotation variant problem in different postures of cow body and then the dominant features (shape distances) are extracted for cow motion classification such as Standing, Lying, and Transition States (Standing-to-Lying and Lying-to-Standing). During the movement of cow motions, the increasing and decreasing trends of shape (Beta features) from cow body are used to classify transition activities of cows. We prepared the datasets by grouping similar motion sequences and tested against with the proposed features. According to experimental results, the proposed system can give the high accuracy with low computational cost in case of detecting and classifying cow motions.
-
Elderly monitoring and action recognition system using stereo depth camera Reviewed
Thi Thi Zin, Ye Htet, Y. Akagi, H. Tamura, K. Kondo, S. Araki
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 316 - 317 2020.10
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
The proposed system used stereo type depth camera by examining the human action recognition and also sleep monitoring in the elderly care center. Different regions of interest (ROI) are extracted using the U-Disparity and V-Disparity maps. The main information used for recognition is 3D human centroid height relative to the floor and percentage of movement from frame differencing for sleep monitoring. The results from the experiments of the proposed method show that this system can detect the person location, sitting or lying and also sleep behaviors effectively.
-
Detection of Estrus in Cattle by using Image Technology and Machine Learning Methods Reviewed
Su Myat Noe,Thi Thi Zin, Pyke Tin, H. Hama
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 320 - 321 2020.10
Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
Detection of estrus in cattle in early phase is especially vital in the era of precision farming. This paper focuses on the detection of estrus in cattle by using image processing techniques and machine learning methods. In doing so, we first utilize an image analysis to investigate some behaviors of cattle in estrus, which is standing when mounted by the other cattle. We then extract some statistical measures based on polyline shape features of detected cattle images and utilize these measures as an input to machine learning algorithms. Specifically, in this paper, we employ the three supervised machine learning methods, which is Support Vector Machine (SVM), Logistic Regression (LR), and Multiple Linear Regression (MLR) classifiers. Some experimental works are performed by using real-life video sequences. The results show promising and capable to detect the behavior of estrus both cost-effectively (only image) and specifically with the detection rate of SVM is 97%, LR is 94%, and MLR is 94%, respectively.
-
Dam Water Overflow Estimation using Time Series Reviewed International coauthorship
Mie Mie Khin, Mie Mie Tin, Thi Thi Zin, Pyke Tin
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 285 - 286 2020.10
Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
This paper will implement the water level estimation of the dam from Myanmar. We use the time series stochastic model to calculate the water level estimation varying on in flow and consumption of dam. This approach is applying probability of Markovian Time Series. This paper based on rainfall and the other factors of dam water storage such as inflow and outflow of dam. This result estimate actual monthly water spread area and shows error is small. This research result also highlight that Markovian Time Series model is one of the best estimate processes for water level estimation.
-
A mobile application for offline handwritten character recognition Reviewed
Thi Thi Zin, Moe Zet Pwint, Shin Thant
2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020 10 - 11 2020.10
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2020 IEEE 9th Global Conference on Consumer Electronics, GCCE 2020
The handwritten character recognition is a computerized system that is able to identify and recognize characters and words written by user. In this paper, we proposed offline handwritten character recognition using deep learning architecture. The Android application of the proposed system is also created by using OpenCV and TensorFlow Lite. The proposed system is aimed to use as a teaching aid in helping kindergarten to primary school level students, especially for practicing their writing and learning. The local handwritten dataset, which includes digit, English alphabet and mathematical symbols that are collected from students, is used for training and testing operations. According to the experimental results, the proposed system is very promising and it will be a useful application for educational environment.
-
A stochastic model for dairy cow body condition scores changes between two successive calving events Reviewed
Thi Thi Zin, Pyke Tin, H. Hama
ICIC Express Letters 14 ( 10 ) 1009 - 1015 2020.10
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
In this paper, we shall propose a stochastic model to investigate and analyze the patterns of dairy cow body condition scores between two successive calving events. Also, a robust Markov chain was introduced and used for stochastic evaluation of body condition score fluctuations from time to time. The study confirmed greater beneficial with increased healthy performance, but a great variation among farms needs to be taken into account. The stochastic model can fully describe the pattern and quantify its characteristics composed of the sum of random variables derived from milk yields, feeding intakes and transition periods in the body energy reserves changes. For this purpose, mathematical modeling techniques can be used to develop decision making systems, in order to achieve optimality of dairy farm management systems. In this aspect, the body condition score plays a key role to make the system successfully carried out. That is to achieve maintaining target score in corresponding periods such as a few weeks after calving, early lactation, mid lactation and dry periods. This concept leads to looking into the dairy cow energy reserves problem of within-the-two successive calving events since the body condition score fluctuation is critical especially at the time of calving, with improvements in production. However, a little has known the statistical and probabilistic tools for relating the body condition score pattern change and milk production, feeding management and animal health during the inter-calving periods. Therefore, we shall formulate the problem of energy reserves in dairy cow body, as a stochastic model of special in which inputs (feed intakes), outputs (mike produced) and the body condition score (energy research storage) are used as random variables. Utilizing a generalized gamma distribution and the univariate normal distribution functions for the marginal and joint distributions of the inputs and outputs in the model, the expected change patterns in body condition scores with respect to time are derived and analyzed. In order to confirm the validity of the proposed method, some simulation results are obtained by using the estimated parameters for inputs and outputs derived from real life dataset. These results show that the proposed approach is well suited to analyze the behaviors of dairy cows associations with body condition scores changing patterns.
-
Classification of People’s Emotions during Natural Disasters
Nann Hwan Khun, Thi Thi Zin, Mitsuhiro YOKOTA, Hninn Aye Thant
宮大工学部紀要 49 ( 49 ) 85 - 90 2020.9
Language：English Publishing type：Research paper (bulletin of university, research institution) Publisher：宮崎大学工学部
Identifying the polarity of sentiments expressed by users during disaster events have been widely researched. At a recent time, social media has been successfully used as a proxy to gauge the impacts of disasters in real-time. With the growing of microblog sites on the Web, people have begun to express their opinions and emotions on a wide variety of topics on Twitter and other similar social services. We proposed a visual emotion analysis framework for natural disasters. The proposed framework consists of two components, emotion analysis modeling and geographic visualization. This emotion analysis modeling is mostly targeted in case of determining the emotions of Twitter users pre, peri and post natural disasters to help first responders for better managing the situations such as mental health of survived victims and fund raising after severe natural disasters. This geographic visualization system can help people for better understanding the changes of emotion reactions along with the duration of natural disasters and mostly interested regions of Twitter users on these natural disasters. In this research, the situations in California Fire which is happened in 2018 November is experimented for emotion analysis because the affected people often show their states and emotions via big data social media environment.
-
Recognition Based Segmentation of Handwritten Alphanumeric Characters Entry on Tablet PC
Myat Thiri Wai, Thi Thi Zin, Mitsuhiro YOKOTA, Khin Than Mya
宮大工学部紀要 49 ( 49 ) 79 - 84 2020.9
Language：English Publishing type：Research paper (bulletin of university, research institution) Publisher：宮崎大学工学部
Portable tablet PC are very useful in relevant industry of this age because tablets are elegant in appearance and convenient to use. Important things are noted on tablet by handwriting easily in respective industry. Recognition of handwritten characters automatically on tablet like human’s brain is also necessary to be more convenient. To split each character of different handwritten styles is very difficult and it is the main challenging of handwritten character recognition. The previous handwritten character segmentation approaches are still continuing in different problems because of different handwritten styles. The combination of sliding windows, region of interest (ROI) box and convolutional neural network (CNN) are used to execute recognition based segmentation (implicit) of handwritten characters. This system is intended to perform both segmentation and recognition of tablet based application input handwritten characters. Handwritten data are collected from 24 members of our laboratory using three tablets PC models to perform the experiments.
-
An Image Technology Approach to Dairy Cow Monitoring System Invited
254 - 255 2020.9
Language：Japanese Publishing type：Research paper (conference, symposium, etc.)
DOI： 10.11527/jceeek.2020.0_254
Other Link： https://www.jstage.jst.go.jp/article/jceeek/2020/0/2020_254/_pdf
-
Important object segmentation and tracking using features and ocr Reviewed International coauthorship
Mie Mie Tin, Mie Mie Khin, Nyein Nyein Myo, Thi Thi Zin and Pyke Tin
ICIC Express Letters, Part B: Applications 11 ( 9 ) 855 - 861 2020.9
Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
Image processing and segmentation support to analyze image and big data in many different research sectors. Object tracking in a surveillance video sequence supports to the research areas. This system processes on the security surveillance camera network and extracts the important object track information. First is the extraction of the key frames from videos. These key frames are used to extract objects on different cameras and from different background areas. To extract key frames from surveillance videos, the system uses RGB color values and the block method with diagonal movement. The important object is segmented from one key frame and the system searches that object a crossing of other key frames. To segment an object, the system uses OTSU’s threshold method. The comparison of an important object and the other segmented objects uses color moment features and computes the similarity value based on the histogram. The system handles all key frames to extract the similar objects. To find the tracking region of that object on different background regions, the system uses time information on key frames in the video networks. To extract time information from video key frames, the system uses character extraction and recognition with the Optical Character Recognition (OCR) method on the gray level images.
-
Robust tracking of cattle using super pixels and local graph cut for monitoring systems Reviewed International journal
Y. Hashimoto, H. Hama, Thi Thi Zin
International Journal of Innovative Computing, Information and Control 16 ( 4 ) 1469 - 1475 2020.8
Authorship：Last author Language：English Publishing type：Research paper (scientific journal) Publisher：International Journal of Innovative Computing, Information and Control
This paper proposes a robust tracking method of Japanese black cattle. Development of a cattle monitoring system using non-contact and non-invasive methods to improve productivity is a strong demand from livestock farmers in aged society. As one of elemental technologies to realize it, we focus on tracking of cattle for detecting estrus behaviors using video camera. The conventional methods like inter-frame difference and background subtraction do not work well under supposed environment. So we propose a new updating method of ROI (Region of Interest) and Scribbles (for foreground and background) according to the movement of the centroid of the extracted cattle region. SP (Super Pixel) and LGC (Local Graph Cut) are adopted for robust cattle region extraction. The tracking without updating soon fails before cattle goes out of frame, but the tracking with the proposed updating has been successfully continued until cattle has gone out. Through the experimental results carried at Sumiyoshi Field attached to Miyazaki University, the effectiveness of the proposed method has been confirmed.
-
Body Condition Score Estimation Based on Regression Analysis Using a 3D Camera Reviewed
Thi Thi Zin, Pann Thinzar Seint, Pyke Tin, Yoichiro Horii and Ikuo Kobayashi
sensors 20 ( 13 ) 2020.7
Language：English Publishing type：Research paper (scientific journal)
The Body Condition Score (BCS) for cows indicates their energy reserves, the scoring for which ranges from very thin to overweight. These measurements are especially useful during calving, as well as early lactation. Achieving a correct BCS helps avoid calving difficulties, losses and other health problems. Although BCS can be rated by experts, it is time-consuming and often inconsistent when performed by different experts. Therefore, the aim of our system is to develop a computerized system to reduce inconsistencies and to provide a time-saving solution. In our proposed system, the automatic body condition scoring system is introduced by using a 3D camera, image processing techniques and regression models. The experimental data were collected on a rotary parlor milking station on a large-scale dairy farm in Japan. The system includes an application platform for automatic image selection as a primary step, which was developed for smart monitoring of individual cows on large-scale farms. Moreover, two analytical models are proposed in two regions of interest (ROI) by extracting 3D surface roughness parameters. By applying the extracted parameters in mathematical equations, the BCS is automatically evaluated based on measurements of model accuracy, with one of the two models achieving a mean absolute percentage error (MAPE) of 3.9%, and a mean absolute error (MAE) of 0.13.
DOI： 10.3390/s20133705
-
Automatic cow location tracking system using ear tag visual analysis Reviewed International journal
Thi Thi Zin, Moe Zet Pwint, Pann Thinzar Seint, Shin Thant, S. Misawa, K. Sumi, K. Yoshida
Sensors (Switzerland) 20 ( 12 ) 1 - 18 2020.6
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：Sensors (Switzerland)
Nowadays, for numerous reasons, smart farming systems focus on the use of image processing technologies and 5G communications. In this paper, we propose a tracking system for individual cows using an ear tag visual analysis. By using ear tags, the farmers can track specific data for individual cows such as body condition score, genetic abnormalities, etc. Specifically, a four-digit identification number is used, so that a farm can accommodate up to 9999 cows. In our proposed system, we develop an individual cow tracker to provide effective management with real-time upgrading enforcement. For this purpose, head detection is first carried out to determine the cow’s position in its related camera view. The head detection process incorporates an object detector called You Only Look Once (YOLO) and is then followed by ear tag detection. The steps involved in ear tag recognition are (1) finding the four-digit area, (2) digit segmentation using an image processing technique, and (3) ear tag recognition using a convolutional neural network (CNN) classifier. Finally, a location searching system for an individual cow is established by entering the ID numbers through the application’s user interface. The proposed searching system was confirmed by performing real-time experiments at a feeding station on a farm at Hokkaido prefecture, Japan. In combination with our decision-making process, the proposed system achieved an accuracy of 100% for head detection, and 92.5% for ear tag digit recognition. The results of using our system are very promising in terms of effectiveness.
DOI： 10.3390/s20123564
-
Body Condition Score Assessment of Depth Image using Artificial Neural Network Reviewed International journal
Thi Thi Zin, Pann Thinzar Seint, Pyke Tin, Y. Horii
ACM International Conference Proceeding Series: ICCMS '20: Proceedings of the 12th International Conference on Computer Modeling and Simulation 33 - 37 2020.6
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：ACM International Conference Proceeding Series
Body Condition Score (BCS) is a visual sign for managing feeding program. It can be described by numeric numbers to estimate available fat reserves and it also allows producers to make better management decisions. Ignoring BCS until it becomes too thin or fat may result in production and animal losses economically. In this paper, we proposed the BCS assessment tool by using depth information from 3D camera. The experimental data were collected in Oita prefecture and BCS scores were taken under the guidance of experts. The information of depth images are used as feature vectors to the input of Artificial Neural Network (ANN) classifier. The proposed system has achieved the good results for classifying two groups of BCS (3.5 and 3.75) with the overall accuracy of 86%.
-
Time to Dairy Cow Calving Event Prediction by Using Time Series Analysis Reviewed International journal
Thi Thi Zin, K. Sumi, Pyke Tin
ACM International Conference Proceeding Series: ICCMS '20: Proceedings of the 12th International Conference on Computer Modeling and Simulation 143 - 146 2020.6
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings) Publisher：ACM International Conference Proceeding Series
In these days the precision dairy farming which is utilization of the Information and Communication Technologies (ICT) has become one of front line research topics in dairy science as well as in data science leading to Agriculture 4.0. An increase in on-farm mortality due to the occurrence calving difficulties with late assistance can cause possible problems not only for animal welfare but also economic losses to the farmers. In this aspect, calving is an extremely important event in the life of a dairy cow. On the other hand, the time around calving is also a critical period since clinical disorders, and calving problems can occur. Calving difficulties are also becoming increasingly common with many dairy cows requiring assistance at the time of calving. To maximize welfare and minimize losses due to calving difficulties, all animals need to be individually monitored to identify any calving difficulties or health problems as early as possible. In addition, it is important to know the exact time of calving event occur so that timely assistance can be made. In this paper, we propose a continuous video monitoring system for time-to calving event investigation based on time series analysis to achieve an accurate calving time prediction. In doing so we have employed three time series models of autoregressive, moving average smoothing. At the same time, we have confirmed the validity of the proposed method by using the real life data experimented on the University Dairy Farm and one of large dairy farms in Japan. The experimental results show that the proposed time series method is promising and can lead to a new prospect in modern precision dairy farming.
-
Image processing technique and hidden markov model for an elderly care monitoring system Reviewed
Swe Nwe Nwe Htun, Thi Thi Zin, Pyke Tin
Journal of Imaging 6 ( 6 ) 49 2020.6
Language：English Publishing type：Research paper (scientific journal) Publisher：Journal of Imaging
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). Advances in image processing technologies have provided more precise views in medical and health care management systems. Among many other topics, this paper focuses on several aspects of video-based monitoring systems for elderly people living independently. Major concerns are patients with chronic diseases and adults with a decline in physical fitness, as well as falling among elderly people, which is a source of life-threatening injuries and a leading cause of death. Therefore, in this paper, we propose a video-vision-based monitoring system using image processing technology and a Hidden Markov Model for differentiating falls from normal states for people. Specifically, the proposed system is composed of four modules: (1) object detection; (2) feature extraction; (3) analysis for differentiating normal states from falls; and (4) a decision-making process using a Hidden Markov Model for sequential states of abnormal and normal. In the object detection module, background and foreground segmentation is performed by applying the Mixture of Gaussians model, and graph cut is applied for foreground refinement. In the feature extraction module, the postures and positions of detected objects are estimated by applying the hybrid features of the virtual grounding point, inclusive of its related area and the aspect ratio of the object. In the analysis module, for differentiating normal, abnormal, or falling states, statistical computations called the moving average and modified difference are conducted, both of which are employed to estimate the points and periods of falls. Then, the local maximum or local minimum and the half width value are determined in the observed modified difference to more precisely estimate the period of a falling state. Finally, the decision-making process is conducted by developing a Hidden Markov Model. The experimental results used the Le2i fall detection dataset, and showed that our proposed system is robust and reliable and has a high detection rate.
-
Usability of Tablet Mobile Devices for Offline Handwritten Character Recognition Reviewed
Thi Thi Zin, T. Otsuzuki
ICIC Express Letters, Part B: Applications 11 ( 6 ) 587 - 593 2020.6
Authorship：Lead author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
In recent years, offline handwritten character recognition has been one of the major topics and challenging research areas in the field of pattern recognition and image processing. Moreover, it is a very problematic research area due to the nature of hand writing styles which can vary from one user to another. On top of that, writings by early learners in formal and non-formal education make the problem more complex and challenging. Generally speaking, the early school age children have diversity in handwriting style, variation in angle, size and shape of characters, making the problems of character recognition more difficult. The number of students has been increasing year by year worldwide. However, due to lack of teachers, many children cannot access high quality education. Therefore, this paper proposes offline character recognition for handwritten characters written on tablet mobile devices. This paper can be considered as an additional supporting system for education of preschool children. This proposed method firstly performs character segmentation process on words acquired from the tablet. In the feature extraction process Histogram of Oriented Gradients (HOG) and Bag of Visual Words (BOVW) are used. Support Vector Machine (SVM) is applied in classification process. Some experimental results are shown to confirm the proposed method.
-
Swe Nwe Nwe Htun, Thi Thi Zin, H. Hama
Applied Sciences (Switzerland) 10 ( 9 ) 3005 2020.5
Language：English Publishing type：Research paper (scientific journal) Publisher：Applied Sciences (Switzerland)
In this paper, an innovative home care video monitoring system for detecting abnormal and normal events is proposed by introducing a virtual grounding point (VGP) concept. To be specific, the proposed system is composed of four main image processing components: (1) visual object detection, (2) feature extraction, (3) abnormal and normal event analysis, and (4) the decision-making process. In the object detection component, background subtraction is first achieved using a specific mixture of Gaussians (MoG) to model the foreground in the form of a low-rank matrix factorization. Then, a theory of graph cut is applied to refine the foreground. In the feature extraction component, the position and posture of the detected person is estimated by using a combination of the virtual grounding point, along with its related centroid, area, and aspect ratios. In analyzing the abnormal and normal events, the moving averages (MA) for the extracted features are calculated. After that, a new curve analysis is computed, specifically using the modified difference (MD). The local maximum (lmax), local minimum (lmin), and half width value (vhw) are determined on the observed curve of the modified difference. In the decision-making component, the support vector machine (SVM) method is applied to detect abnormal and normal events. In addition, a new concept called period detection (PD) is proposed to robustly detect the abnormal events. The experimental results were obtained using the Le2i fall detection dataset to confirm the reliability of the proposed method, and that it achieved a high detection rate.
DOI： 10.3390/app10093005