Paddy insect pest identification using artificial intelligence and machine learning

Lai, Zhi Yong (2020) Paddy insect pest identification using artificial intelligence and machine learning. [Project Paper] (Submitted)

Text FK 2020 20.pdf Download (2MB)

Abstract

The trend of paddy harvested area in Malaysia has a -0.2% growth rate where the government has to import about 0.9 million metric tonnes (MT) of rice from neighboring countries per year. The government strives to achieve 100% self-sufficiency level of rice in 2020. Pest has been a critical challenge in paddy plantation as it can cause up to 37% loss in yield. This project developed an insect pest detection model that can detect the species of insect pest as well as counting the insects according to their species using Faster Recurrent Convolutional Neural Network (Faster R-CNN). The insects were captured at MARDI with a sticky pad light trap, and the images of the insects were taken using a smartphone camera which was set at a shutter speed of 1/50 second, lens aperture of f/2 and a focal length, f = 3.5mm. The images were taken in a black box with 6 LED light bulb. The model was trained with 1460 RGB images of 3 classes of insects including Zig Zag brown planthopper (Zig Zag BPH), green leafhopper (GLH) and a class for a mixture of other insects. The model was trained for 23200 iterations and the accuracy of classification of Zig Zag BPH, GLH and others were 0.98, 1 and 0.97 respectively, while the F1 score for the respective species were 0.92, 1 and 0.98. The model was able to achieve a mean average precision (mAP) of 0.93 and an average recall (AR) of 0.60

Item Type:	Project Paper
Faculty:	Faculty of Engineering
Depositing User:	Ms. AZLINA ZAINAL ABIDIN
Date Deposited:	11 Nov 2022 03:57
Last Modified:	11 Nov 2022 07:07
URI:	http://psaspb.upm.edu.my/id/eprint/565

Actions (login required)

View Item