| Autor: |
Raja R., Vinston, V., Adithya, Hollioake, Flintoff A., N., Krishna Kumar, P. L., Kirran, Dhanalakshmi, G. |
| Jazyk: |
angličtina |
| Rok vydání: |
2023 |
| Předmět: |
|
| Zdroj: |
International Journal of Intelligent Systems and Applications in Engineering; Vol. 11 No. 3 (2023): (Articles in Press); 01-12 |
| ISSN: |
2147-6799 |
| Popis: |
The aim of this research paper is to use deep learning and R-CNN features with open cv tools to detect and recognize different fish species from underwater images. The RCNN method uses a selective search algorithm to extract the top 2000 region proposals from millions of Regions of Interest (RoI) proposals in an image, which is then inputted to a CNN model for further analysis. The deep learning approach used for fish recognition achieved 85% accuracy. The focus of this study is on fish recognition in a natural lake to aid in preserving the original environment. To achieve this, the paper proposes a scheme for segmenting fish images and measuring their morphological features using R-CNN. The process starts with acquiring fish body images using a homemade image acquisition device, preprocessing and labeling the images, and then training the R-CNN model with the labeled images. The fish images are segmented using the trained model, enabling the extraction of morphological characteristics that serve as indicators of the fish. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
