Ramdan, Yoki Andriawan and Setiawan, Widi and Sofyan, Edi (2021) Development of Embedded Image Processing to Classify, Determine and Control Nutrient Deficiency of Plantation in Hydroponic System. Masters thesis, Swiss German University.

Full text not available from this repository.

Abstract

One of required technology to growth Hydroponic system is how to quickly detect nutrient deficiencies in plants. In the traditional way, checking for nutrient deficiencies is done manually, using destructive method by taking sampling of the leaf. This requires a lot of time and effort, especially for large areas. In this thesis, Deep Convolutional Neural Network is used to diagnose Fe deficiency of Pak Choy. We compare Inception-ResnetV2 and MobileNetV2 architecture using transfer learning method with the fine-tuning model was carried out to train dataset which consist of 249 images for training and 28 images for testing. The result show that best accuracy from the training is achieve 98% for 2000 epochs using MobileNetV2 and was finally selected to deploy to the Raspberry PI 4B, with test in live capture image show accuracy 81%. The average prediction time process in live capture image about 3 seconds. The Fe Deficiency detection algorithm integrated with nutrient dosing control system able giving the feedback from Fe Deficiency detection to run the dosing pump A2 for Fe nutrient. The dosing control show good result, which is based on time calculation, able to run 5ml dosage in 10s with speed setting of dosing pumps 36 rpm.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Deep Learning, Deficiency Detection, Dosing Control, Embedded System, Nutrient Deficiency
Subjects:	?? SB126.5 ?? ?? TJ ?? ?? TJ163.12 ??
Divisions:	?? sch_che ??
Depositing User:	Users 2 not found.
Date Deposited:	17 Jan 2022 08:30
Last Modified:	17 Jan 2022 08:30
URI:	http://repository.sgu.ac.id/id/eprint/2325

Actions (login required)

View Item