Agriculture is currently one of Ethiopia's most important economic sectors. Numerous cattle, both domestic and foreign, are being raised for a variety of uses. The current state of the identification of cattle breed in Ethiopia still in manual, and only the domain experts try to identify every cattle breed type. The identification of cattle breed type are depends on the skills and experience of experts in the domain area, and this can lead to cause of error, inaccurate, time-consuming, and tedious. Therefore, it is necessary to design a Identification model that assists the selection mechanism for cattle breeds.The primary aim of this study is to develop a cattle breed identification model using a deep learning approach to identify the cattle breeds. In this study, we used 2,500 images that include different breeding of cattle namely Jersey, Holstein-Friesian, Minamir, Brown Swiss and Local. We partitioned the total image into 70% for training and 30% for testing. We applied Median, Gabor and Gaussian filter noise removal techniques to remove the noise that occurred during image acquisition. To increase the number of image, we also applied an image augmentation technique. We fed the processed image to the Cattle Breed Identification Model(CBIM), VGG16, and Resnet50. We evaluate the performance of those models individually, using the accuracy, confusion matrix, recall, f1-score, and precision in different conditions for the Median, Gaussian, and Gabor-filtered images separately.The performance results for the Cattle Breed Identification model for the Median, Gaussian, and Gabor filtered images are 93%, 91%, and 81%, respectively. We get an accuracy of 82%, 80%, and 66%, respectively, using VGG16. In addition, for the Median, Gaussian-applied, and Gabor-filtered images, the Resnet50 achieves accuracy of 64%, 52%, and 58%, respectively. Then we get the accuracy of 93%, 82%, and 64%, respectively, for the Cattle Breed Identification, VGG16, and Resnet50 on the Median-filtered images. The overall performance of the model is 93% using the Cattle Breed identification model. This identification model will play a great role while, deploy for different agriculture research institutes. It will also advantages for different researchers who interested to work in this domain area.

Agriculture is currently one of Ethiopia's most important economic sectors. Numerous cattle, both domestic and foreign, are being raised for a variety of uses. The current state of the identification of cattle breed in Ethiopia still in manual, and only the domain experts try to identify every cattle breed type. The identification of cattle breed type are depends on the skills and experience of experts in the domain area, and this can lead to cause of error, inaccurate, time-consuming, and tedious. Therefore, it is necessary to design a Identification model that assists the selection mechanism for cattle breeds.The primary aim of this study is to develop a cattle breed identification model using a deep learning approach to identify the cattle breeds. In this study, we used 2,500 images that include different breeding of cattle namely Jersey, Holstein-Friesian, Minamir, Brown Swiss and Local. We partitioned the total image into 70% for training and 30% for testing. We applied Median, Gabor and Gaussian filter noise removal techniques to remove the noise that occurred during image acquisition. To increase the number of image, we also applied an image augmentation technique. We fed the processed image to the Cattle Breed Identification Model(CBIM), VGG16, and Resnet50. We evaluate the performance of those models individually, using the accuracy, confusion matrix, recall, f1-score, and precision in different conditions for the Median, Gaussian, and Gabor-filtered images separately.The performance results for the Cattle Breed Identification model for the Median, Gaussian, and Gabor filtered images are 93%, 91%, and 81%, respectively. We get an accuracy of 82%, 80%, and 66%, respectively, using VGG16. In addition, for the Median, Gaussian-applied, and Gabor-filtered images, the Resnet50 achieves accuracy of 64%, 52%, and 58%, respectively. Then we get the accuracy of 93%, 82%, and 64%, respectively, for the Cattle Breed Identification, VGG16, and Resnet50 on the Median-filtered images. The overall performance of the model is 93% using the Cattle Breed identification model. This identification model will play a great role while, deploy for different agriculture research institutes. It will also advantages for different researchers who interested to work in this domain area.

DSpace Repository

UOG Instituional Repository Home
→
College of Natural and Computational Sciences
→
Physics
→
Physics
→
View Item

Agriculture is currently one of Ethiopia's most important economic sectors. Numerous cattle, both domestic and foreign, are being raised for a variety of uses. The current state of the identification of cattle breed in Ethiopia still in manual, and only the domain experts try to identify every cattle breed type. The identification of cattle breed type are depends on the skills and experience of experts in the domain area, and this can lead to cause of error, inaccurate, time-consuming, and tedious. Therefore, it is necessary to design a Identification model that assists the selection mechanism for cattle breeds.The primary aim of this study is to develop a cattle breed identification model using a deep learning approach to identify the cattle breeds. In this study, we used 2,500 images that include different breeding of cattle namely Jersey, Holstein-Friesian, Minamir, Brown Swiss and Local. We partitioned the total image into 70% for training and 30% for testing. We applied Median, Gabor and Gaussian filter noise removal techniques to remove the noise that occurred during image acquisition. To increase the number of image, we also applied an image augmentation technique. We fed the processed image to the Cattle Breed Identification Model(CBIM), VGG16, and Resnet50. We evaluate the performance of those models individually, using the accuracy, confusion matrix, recall, f1-score, and precision in different conditions for the Median, Gaussian, and Gabor-filtered images separately.The performance results for the Cattle Breed Identification model for the Median, Gaussian, and Gabor filtered images are 93%, 91%, and 81%, respectively. We get an accuracy of 82%, 80%, and 66%, respectively, using VGG16. In addition, for the Median, Gaussian-applied, and Gabor-filtered images, the Resnet50 achieves accuracy of 64%, 52%, and 58%, respectively. Then we get the accuracy of 93%, 82%, and 64%, respectively, for the Cattle Breed Identification, VGG16, and Resnet50 on the Median-filtered images. The overall performance of the model is 93% using the Cattle Breed identification model. This identification model will play a great role while, deploy for different agriculture research institutes. It will also advantages for different researchers who interested to work in this domain area.

Worku, Tadesse

URI: http://hdl.handle.net/123456789/8752

Date: 2024-09

Abstract:

In this project, employing the master equation we have obtained the quantum Langevin equation and the equation of time evolution for the expectation values of the cavity mode atomic operators. With the aid of the solutions of these equations, we have calculated the correlation properties for the cavity mode operators. Using the obtained correlation properties for cavity operators and the large time approximation scheme, we determined the master equation, the quantum lengevin equation and the correlation for the cavity operator and We have found that the effect of this interaction is to decrease the quadrature squeezing and the mean photon number of the two-mode cavity light.

Show full item record