Browsing by Author "SIDDIK, Othman"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item AUTOMATIC SPIRULINA DETECTION USING IMAGE PROCESSING TECHNIQUES(2022-02-14) SIDDIK, Othman; BOSTAN, AtilaIn this thesis, a study on automatic detection of spirulina is presented. Spirulina is an algae microorganism with 4 species which are quite useful for the determination and monitoring of water quality. Thesis contribution is to develop an automatic process for helping the diagnosis Spirulina in water, most of the Spirulina can be diagnosed by the size and shape from microscopic images, all algae detection that has to be diagnosed in a fast and accurate way is very critical for the water quality, manual methods are used to detect spirulina. This can give rise to inaccurate results. It is also very tedious effort to detect algae within water microscopic images. Automatic detection of spirulina is a challenging task due to factors such as change in size and shape with climatic changes, growth periods and water contamination. Nowadays, the automated detection of spirulina is one of the most fervent topics in applied biology. On the other hand, Deep-Learning and Convolutional Neural Networks (CNN) is yielding better results and is a judiciously used technique for image classification and for a variety of problems. This thesis introduces CNN into the automated spirulina detection problem in order to demonstrate whether it would succeed in solving the spirulina detection problem. A comprehensive dataset was specifically prepared using an artificial image generation method out of original images that are collected from rivers and lakes in Turkey. In this study, a spirulina image data set was prepared using a customized technique for artificial image generation. Consequently, a dataset covering different illumination conditions was computationally augmented to 1000 sample images. Original images were collected from rivers and lakes in Turkey. In this thesis, the background to the spirulina detection problem, the methodology used in the study and the results of image processing and feature extraction methods to locate and extract spirulina in a microscopic image are reported. Initially, the RGB image format with morphological operations were employed to detect spirulina in a microscopic image. As a result with a rate of 84% accuracy detection was observed. Afterwards, three different methods were experimented with for comparison purposes. The methods and their relative detection success rates were observed as follows: SURF 63%, FAST feature detection 67%, CNN 99% result accuracy rate, consequently, some future work is also suggested to improve the study further. In this thesis, we introduced CNNs into the automated spirulina detection problem. A CNN method used to solve 4 class spirulina detection problem. Observed results were discussed and compared with those of previous studies. To the best of our knowledge and survey results on the literature, this is the first study to employ CNNs in the automated spirulina detection problem.