Underwater Object Detection
Underwater Object Detection
Underwater Object Detection
http://doi.org/10.22214/ijraset.2020.5344
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 8 Issue V May 2020- Available at www.ijraset.com
Abstract: Detecting underwater objects is challenged by various kinds of aspect ratios, object size,material colour, cluttered
backgrounds, and in particular,not defined orientations. In our paper, we are using a Probabilistic Neural Network (PNN)
features from layers which are combined to perform orientation robust aerial underwater object detection. We explore the
essential characteristics of PNN as well as corelate the extracted features to the principle of disengaging feature learning. An
image segmentation based approach is used to localize ROIs of different aspect ratios, and furtherly ROIs are classified into
positives or negative using a DCNN features. On inquiring the two datasets collected from Google Earth, we illustrate that the
proposed aerial underwater object detection approach is simple and easy process. Fast and robust underwater object detection in
aerial images is potentially applicable in traffic surveillance, emergency, remote sensing and large scale image content analysis.
Keywords: Image processing, Deep Convolutional Neural Network(DCNN), Discrete Wavelet Transform, Underwater Object
Detection, GLCM features, GRNN.
I. INTRODUCTION
In an image, the object indentification uses image forecasting techniques by removing the noise, followed by the regions, locating
lines and areas with some textures using feature extraction. This identification of objects in an image suffers disadvantage such as
different angles and different lighting. The human visual system does these tasks unconsciously to equalise human performance for
a computer, a skillfull program and great processing power is required. In computer vision, recognizing objects is a long time goal.
The reason for this is,in real time the variations of matter instances belonging to the same type appear to be similar in an image.
Hence the paper briefs two goals, recognition and classification of images/object detection. The primary goal in this paper is an
approach for object class recognition that employs edge information only. It is represented by very simple and generic shape
primitives of curved shapes, especially ellipses and line segments. Here, each combination is a two-layer abstraction of primitives:
pairs of primitives (termed shape tokens) at the first layer, and a learned number of shape tokens at the second layer.
A. Advantages
1) Accurately classify large number of coral plants
2) No need for large training datasets
A. Pre-Processor
Image Pre-processing is operations that are performed with images at the abstraction of lowest levels. Its input and output are
intensity images. The ultimate aim of pre-processing is suppressing unwanted distortions and enhances some image features which
is important for further processing and improving the data of an image . Image restoration is defined as finding a corrupted/noisy
image and estimating the clean original image. Corruption may come in many forms such as noise, camera misfocus,motion blur
.and More advanced image processing techniques must be applied to recover the object. Example: deconvolution of image
restoration method. It is capable of increasing contrast , resolution,specially in the axial direction removing noise
B. DWT
The DWT gives a scattered representation for many natural signals. a subset of DWT coefficients are used for capturing the
important features of natural signals which is much smaller than the original singals. This helps in compressing the signal. With the
help of DWT, we always have the same number of coefficients as the original signal, but many of the coefficients we obtained may
be close to zero.As a result, we can send those coefficients for maintaining a high-quality signal approximation. Basically DWT is
used for orthogonal transform and not for shift invariant. The operation of discrete filter banks are done by discrete wavelet
transforms. This provides the perfect reconstruction of the signal upon inversion which helps us to take the discrete wavelet
transform of a signal and then using the coefficients we can synthesize an exact reproduction of the signal to numerical precision
C. GLCM Features
The gray-level co-occurrence matrix (GLCM) can be obtained by gray co matrix function. This is for calculating how often a pixel
with the intensity (gray-level) value i occurs in a value j which has a specific spatial relationship to a pixel . By default, The pixel of
interest and the pixel to its immediate right (horizontally adjacent) is said to be spatial relationship, Though we can specify other
spatial relationships between the pixels (ie) i and j. The resultant GLCM with element (i,j) is simply the sum of the number of times
that the pixel with value i occurred in the provided with value j spatial relationship to a pixel in the input image.
These statistics information provide us about the texture of an image. Statistic such a as Energy ,Correlation, , Homogeneity,
Contrast gives information about image
D. Neural Network
Neural Network (NN) and General Regression Neural Networks (GRNN) have similar architectures, but there is a fundamental
difference: networks perform classification where the target variable is categorical, whereas for general regression neural networks
perform regression continuous target variable is obtained. If we select a NN/GRNN network, selection of the correct type of
network based on the type of target variable will be done automatically with the help of DTREG.
Figure 4. Architecture of NN
V. HARDWARE REQUIREMENTS
A. Arduino
ARDUINO(ATMEGA) is an open source hardware device used for interfacing the all sensors and electronic devices used to
perform a specific task. Any unique task can be completed by loading a set of instructions(program code) via a serial connection
from the computer to arduino board . Arduino has 2 parts one is physical programmable circuit board and another one is a piece of
software or IDE. This arduino software works on the computer, which is used to write and upload computer code to the physical
board.
The Arduino boards which we are using is programmed via Universal Serial Bus (USB)which can be implemented using USB-to-
serial adapter chips such as the FTDI FT232. Some boards, which are manufactured before such as later-model Uno boards, replace
the FTDI chip with a separate AVR chip that holds the USB-to-serial firmware.
REFERENCES
[1] Forest Rohwer, Merry Youle, and Derek Vosten, Coral reefs in the microbial seas, vol. 1, Plaid Press United States, 2010.
[2] MR Patterson and NJ Relles, “Autonomous underwater vehicles resurvey bonaire: a new tool for coral reef management,” in Proceedings of the 11th
International Coral Reef Symposium, 2008, pp. 539–543.
[3] Jean C Kenyon, Russell E Brainard, Ronald K Hoeke, Frank A Parrish, and Casey B Wilkinson, “Towed-diver surveys, a method for mesoscale spatial
assessment of benthic reef habi- tat: a case study at midway atoll in the hawaiian archipelago,” Coastal Management, vol. 34, no. 3, pp. 339–349, 2006.
[4] Oscar Beijbom, Peter J Edmunds, David Kline, B Greg Mitchell, David Kriegman, et al., “Automated annotation of coral reef survey images,” in Computer
Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. IEEE, 2012, pp. 1170–1177.
[5] B Boser Le Cun, John S Denker, D Henderson, Richard E Howard, W Hubbard, and Lawrence D Jackel, “Handwritten digit recognition with a back-
propagation network,” in Advances in neural information processing systems. Citeseer, 1990.