Development of a coin counting system by using machine vision

Liew, Lee Hao (2024) Development of a coin counting system by using machine vision. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full text) Development of a coin counting system by using machine vision.pdf - Submitted Version Download (11MB)

Abstract

This research focuses on developing an automated coin counting solution using machine vision technology to improve upon existing coin counting systems, particularly targeting small businesses facing inefficient and costly coin counting methods. The objectives of this study is to create a portable coin counting system. Also, by using image processing techniques with Python OpenCV, the research will also delve into the analysis of speed and accuracy of the coin counting system. The methodology involves designing a portable casing, preprocessing coins using greyscale conversion and various image enhancement techniques, detecting coin contours, differentiating coins based on sizes and colours and displaying the final results. Then, the experiment was done across 5 set of coin values and the average time taken to calculate the coins as well as average accuracy in terms of coin detection and calculation was obtained. Besides, the mean error as well as mean absolute error of the accuracy under different conditions were also explored. Tests conducted show promising results, with the system achieving an average calculation time of less than 0.013 s for different coin sets as well as 100% accuracy under optimal conditions in which medium brightness, light intensity of warm yellow or red tone as well as area of image detection after cropped should be used. However, the system has limitations, including low camera resolution and complex lighting issues. To address these, recommendations are made to improve the system, such as integrating a high-resolution camera, using high-speed data transfer technology, employing spiking neural networks, and utilizing a diffuse or dome light source. Future research could also focus on including more coin denominations and detecting counterfeit coins. Overall, this research had achieved all its objectives and is able to contribute to the development of more reliable and efficient coin recognition systems.

Actions (login required)

View Item