Image processing and analysis techniques for estimating weight of mangoes

Ngahadi, Nur Azreen Azwa (2021) Image processing and analysis techniques for estimating weight of mangoes. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Image processing and analysis techniques for estimating weight of mangoes.pdf - Submitted Version Download (351kB)
	Text (Full text) Image processing and analysis techniques for estimating weight of mangoes.pdf - Submitted Version Restricted to Repository staff only Download (1MB)

Abstract

In today’s scenario, image processing has played a major role which frequently implemented in applications such as image sharpening and restoration, colour processing, quality grading and inspections. This project was focused on image sharpening and restoration which is referred to as process images that been captured from the image acquisition setup. For quality grading, it is widely used in agriculture applications such as fruit quality in terms of weight, volume, colour, size and shape. The problem of quality grading on the farm is the grading process being carried out manually. The results of the manual grading procedure are typically unreliable, timeconsuming and prone to mistakes, given the different skill levels and experience of human workers. The project planned to carry out an image processing algorithm and analysis techniques for measuring the weight of mangoes. Depending on the applications of the image processing algorithm, several criteria had to be taken into consideration such as the size of the mango image in terms of high resolution or the speed of the performance in processing the image. Used for processing and analysis of pictures taken from the image capture equipment is Matrix Laboratory (MATLAB) software. The image processing algorithm was including image enhancement, image adjustment, image thresholding and features extraction. The number of pixels from the image was then tallied in the mango region. In order to analyse the relationship between actual mango and mango pixels, the linear regression model was built. The correlation coefficient, R2 value for the method developed is 0.98. The performance of the method was evaluated in terms of accuracy and error to know the method is valid to use or not. The weight estimation method obtained the Root Mean Square Error (RMSE) value of 2.29 % and the accuracy is 97.71%. For the size classification, the accuracy is 100% by using the Gaussian Naive Bayes classifier learner with the training time is 0.1 second.

Actions (login required)

View Item