Automated Grading of Edible Birds Nest with Bat Algorithm

Gan, Jack En (2019) Automated Grading of Edible Birds Nest with Bat Algorithm. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Gan Jack En.pdf Restricted to Registered users only Download (3MB)

Abstract

Edible Bird's Nest (EBN) is one of the indulgences in the Chinese cuisine. Traditionally, EBN was accepted to have numerous health advantages. Malaysia is the third biggest producer of Edible Bird Nest (EBN). The most common way of grading an EBN is by using human observation. If a system can be trained to carry out the grading process with high accuracy, it would help in reducing the cost of hiring the experts. The EBN was graded based on their size, shapes (or curvature), impurities on the nest and color of the nest. The EBN was categorized into 4 grades: Grade AA has the highest quality followed by Grade A, Grade B, and Grade C. Bat Algorithm (BA), Bat Algorithm and K-Means (BAKM) and K-Means and Bat Algorithm (KMBA) will be used as the classifier to grade the EBN. The results obtained by each classifier will be compared and evaluate their performance. New unprocessed EBN image will undergo image processing to extract its features data (Size, Curvature, Impurities and Color). The EBN grade used for the new EBN datasets was Grade A, Grade B and Grade C. Image processing techniques such as image morphologies, edge detection, HSV color model and thresholding were used in the program. The data will then be normalized and save as datasets for the next process. Overall BAKM and KMBA was concluded to be have the best performance depending on which datasets was used. The result of classification old EBN datasets by BAKM gives a highest average accuracy of 85.21% while KMBA gives a highest average accuracy of 85.6%. Decimal normalization method of old EBN gives the best result among 3 normalization method. Various recommendations and improvements were suggested after analyzed the new EBN image and results to improve the accuracy of the features dataset extracted from the EBN images in the future.