Marine science Thesis Proposal, UTM, Malaysia Fill a knowledge gap in comparing filter algorithms for classifying marine objects from UAV data

Justification of the Study

The purpose of this work is to fill a knowledge gap in comparing filter algorithms for classifying marine objects from UAV data. Building accessible visual perception systems are essential for marine surveillance, which requires highly effective and accurate detection of marine objects. Therefore, there is a need to aggregate high-performance deep learning-based methods and high-quality datasets relevant to the marine environment. Data has several properties that make it difficult to process and extract information. These characteristics are resolution non-uniformity, intensity non-uniformity, speckle noise due to mutual interference of sampled acoustic echoes, acoustic shadowing, reverberation, and sonar image because the acoustic image is a 2D horizontal projection of the observation environment. design concept. This fact causes an ambiguity problem because equidistant objects at different heights map to the same position in the acoustic image. Previous studies have relied on radio-controlled (RC) unmanned aerial vehicles (UAVs), also known as RPVs, to collect meaningful data in the atmosphere (Maldonado-Ramirez, 2005). As part of their current research, scientists have developed a UAV system that improves position and altitude accuracy and collects more precise data. Unmanned Aerial Vehicles (UAVs) are widely used in civilian infrastructure due to their low maintenance cost, ease of deployment, hovering performance, and excellent maneuverability. UAVs can capture images faster and more accurately than satellite imagery, allowing for faster assessments. One of the main reasons for their development is human limitations. Not having people has several benefits, including increased maneuverability, better stealth capabilities, and nearly unlimited duration. In most cases, they are cheaper than conventional aircraft.

2.1.2   Low Pass Filter

A  low pass filter is a filter that reduces high frequencies while leaving low frequencies intact (Hardik. P., 2011). It is mainly used for smoothing and blurring images. There are three different types of filters: Ideal, Butterworth, and Gaussian. The simplest low pass filter is the ideal low pass filter. Suppresses all high frequencies compared to the cutoff frequency r0 and passes low frequencies unchanged. The ILPF transfer equation is defined as :

Where:

M is the number of rows in the image

N is the number of columns in an image

(M/2, N/2) is the center of frequency rectangle

Is the transform point (u, v) from the center.

Butterworth low pass filters keep the frequencies within the radius D0 and the outer rejection value introduces a gradual transition from 1 to 0 to reduce ringing artifacts. Note that the blur transitions smoothly as the cutoff frequency increases. No ringing is seen in any image processing with this particular BLPF. BLPF transfer equation is :

Where:

M is the number of rows in the image

N is the number of columns in the image

(M/2, N/2) is the center of frequency rectangle

Is the transform point (u,v) from the center  N is the transfer order.

Gaussian low pass filters attenuate high-frequency signals. Gaussian filtering does not clip the frequency coefficients abruptly but instead uses a smoother cutting process. GLPF transfer equation is :

Where:

M is the number of rows in the image

N is the number of columns in the image

(M/2, N/2) is the center of the frequency rectangle.

Answer

Request Answer of this Assignment