Leaf blast spot detection method based on Linknet

Chen Junshen; Gong Meiling; Dong Wang; Xiaoxia Zhao; Min Song; Huimin Guo

doi:10.1117/12.2553085

31 January 2020 Leaf blast spot detection method based on Linknet

Chen Junshen, Gong Meiling, Dong Wang, Xiaoxia Zhao, Min Song, Huimin Guo

Proceedings Volume 11427, Second Target Recognition and Artificial Intelligence Summit Forum; 1142742 (2020) https://doi.org/10.1117/12.2553085
Event: Second Target Recognition and Artificial Intelligence Summit Forum, 2019, Changchun, China

Abstract

This paper aims to solve the problem of automatic detection of rice leaf lesions in natural scenes using deep learning techniques. In this paper, the Linknet full convolutional network was built to train the segmentation model. The network compensates the lost spatial information in the feature extraction process through the short connection structure between downsampling and corresponding upsampling. The model takes rice canopy RGB image as input and then output binarized lesion segmentation image. Then considered with the distribution characteristics of lesion spots, the loss function of the origin model was replaced with Focal loss function, which further improved the segmentation accuracy of the model. The average precision and recall have respectively achieved 98.55% and 98.64% on validate data set, and the average false positive rate has reduced to 1.36%, which has a better segmentation performance. It creates a good precondition for automatic identification of leaves diseases.

Citation Download Citation

Chen Junshen, Gong Meiling, Dong Wang, Xiaoxia Zhao, Min Song, and Huimin Guo "Leaf blast spot detection method based on Linknet", Proc. SPIE 11427, Second Target Recognition and Artificial Intelligence Summit Forum, 1142742 (31 January 2020); https://doi.org/10.1117/12.2553085

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available