An Innovative Method for Detection of Insect Based on Mask-R-CNN Approach

Seventy percent of India's labor force works in agriculture, according to a recent survey. Pests and illnesses cause both qualitative and quantitative losses in crop production. While automatic in-field pest detection using a computer vision approach is an important part of modern intelligent a...

Full description

Saved in:

Bibliographic Details
Published in:	2023 7th International Conference on Electronics, Communication and Aerospace Technology (ICECA) pp. 559 - 564
Main Authors:	Geetha, B.T., Jiwatode, Vilas Ramkrushna, Raut, Ranjit Raosaheb, Hussan, M.I. Thariq, Tiwari, Mohit, Dobhal, Dinesh Chandra
Format:	Conference Proceeding
Language:	English
Published:	IEEE 22-11-2023
Subjects:	Convolutional Neural Network (CNN) Convolutional neural networks Feature extraction Gray Level Co-occurrence Matrix (GLCM) Image preprocessing Image segmentation Insects Manuals Object segmentation Region Proposal Network (RPN)
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Seventy percent of India's labor force works in agriculture, according to a recent survey. Pests and illnesses cause both qualitative and quantitative losses in crop production. While automatic in-field pest detection using a computer vision approach is an important part of modern intelligent agriculture, there are still significant challenges to be overcome. These include the complexity of the natural environment, the detection of tiny size pests, and the classification into several classes of pests. The proposed method consists of four stages: image preprocessing, segmentation, GLCM, and analysis. Model training and feature extraction. Image preprocessing techniques allow for the processing of a still image and the creation of an image improvement approach. The segmentation procedure makes use of grayscale insert images. A feature extraction tool that makes good use of the Gray level co-occurrence matrix. The models are then trained via Mask R-CNN, after information gain has been used to choose relevant features. CNN and RPN, two of the most common substitutes, are both outperformed by the proposed strategy.
DOI:	10.1109/ICECA58529.2023.10395729