Publication: Smart iot based monitoring system for mushroom farming and transfer learning to classify the grading of grey oyster mushroom
Loading...
Date
2023-07
Authors
Teng, Xing Sheng
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This project introduces an application that can continuously monitor crucial parameters, including temperature and humidity, which have a significant impact on
mushroom growth. The study focused on growing grey oyster mushrooms, which required humidity levels between 80% to 90% and temperatures below 30°C.
Traditional methods of managing these elements and classifying the grade of grey oyster mushroom require substantial human effort and are time-consuming. Therefore, a smart Internet of Things (IoT) based mushroom monitoring system with transfer learning to classify the grading of oyster mushroom is proposed, utilizing ESP8266 as the control hub and DHT22 as the sensor module. The sensor system is linked to Blynk application to enable IoT features. The data collected by the sensor is displayed on Blynk webpage and its mobile application. Farmers can access and monitor real-time data using a mobile phone, laptop or desktop. Besides, this project also implement an alert system to notify the user when the temperature of the farm is higher than 30°C or the humidity level is less than 80%. When the alert system is triggered, the email notification and push notification will also send to the farmer. Furthermore, this project also proposed an application which use transfer learning to classify the grading of grey oyster mushroom. A pretrained neural network, ResNet-50 with validation accuracy of 95.37% is used to classify the grading of mushroom using MATLAB programming software. The data set of the grey oyster mushroom consisted of total 1612 images with grade fresh and defected. The number of training images are 943 while the number of validation images are 669. Testing process is done by using 10 testing images and ResNet-50 correctly identified the grading of the mushrooms with testing accuracy of 100%.