A Feature Learning and Object Recognition Framework for Underwater Fish Images with Segmentation
Journal Title: Elysium Journal of Engineering Research and Management - Year 2016, Vol 4, Issue 3
Abstract
Live fish recognition is one of the most decisive elements of fisheries analysis where the massive amount of data is promptly assimilated .Diverse from wide-ranging scenarios, experiments to underwater image recognition are forwarded by poor image quality, uncontrolled objects and environment, and efforts in obtaining demonstrative illustrations. In addition, most existing feature extraction techniques are caught up from automation due to human observation. Hence, we propose an underwater fish recognition framework that comprises of an entirely unsupervised feature learning procedure and an error-resilient classifier. Object parts are modified based on saliency and slackening cataloging to match objective parts appropriately. A non-rigid typical part is then utilized based on fitness, partition, and refinement measures. For the classifier, an unsupervised clustering methodology yields a binary class order, where every individual node acts as a classifier. The main objective of this process is to appraise the basic beliefs of the efficient processes usually used to analyze type of fish. Experimentations illustrate that the proposed framework attains high accurateness on both public and self-collected underwater fish images with high uncertainty and class imbalance. Also to recover the effectiveness of the progression when related with the other existing process.
Authors and Affiliations
Letcy Fernando O, Rexy J
Keywords
Related Articles
- EP ID EP365648
- DOI -
- Views 106
- Downloads 0
Electrocardiogram Signal Modeling Using Adaptive Framework Based SVM Classification Method
Automatic electrocardiogram (ECG) signal classification plays a significant role in the clinical applications, to overcome the problems occur during manual annotation of the ECG recordings. The ECG beat m...
Implementation of Weight Monitoring Using RFID and Load Cell
An embedded system is the combination of both software and hardware components. Real Time Operating System (RTOS), hardware, application software are the components of embedded system. The embedded system is widely used...
REAL TIME SECTIONALIZATION OF ENHANCED SHARPNESS VIDEO USING FPGA
Background Identification is a general feature in many video privilege systems. Gaussian Mixture Models (GMM) is one of the popular fashionable and winning approaches to complete Background identification...
PREDICTION IN DATA MINING COMPARATIVE STUDY OF HEART DISEASE
Data mining is the process of finding useful and relevant information from the databases. There are several types of data mining techniques are available. Association Rule, Classification, Neur...
SECURE PRIVACY PRESERVING TECHNIQUES IN SMART PHONES : A REVIEW
The context aware applications are blooming with increasing the popularity of smartphones with sensors, these sensor applications arises the privacy is the major challenging issue because users...