אסיף מאגר המחקר החקלאי

Robotic Disease Detection in Greenhouses: Combined Detection of Powdery Mildew and Tomato Spotted Wilt Virus

Year:

2016

Source of publication :

Authors :

Volume :

1

Co-Authors:

Schor, N., Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, Beer-Sheva, Israel, Institute of Agricultural Engineering, Agricultural Research Organization, Beit Dagan, Israel
Bechar, A., Institute of Agricultural Engineering, Agricultural Research Organization, Beit Dagan, Israel
Ignat, T., Institute of Agricultural Engineering, Agricultural Research Organization, Beit Dagan, Israel
Dombrovsky, A., Institute of Plant Protection, Agricultural Research Organization, Beit Dagan, Israel
Elad, Y., Institute of Plant Protection, Agricultural Research Organization, Beit Dagan, Israel
Berman, S., Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, Beer-Sheva, Israel

Facilitators :

Link :

Abstract:

Robotic systems for disease detection in greenhouses are expected to improve disease control, increase yield, and reduce pesticide application. We present a robotic detection system for combined detection of two major threats of greenhouse bell peppers: Powdery mildew (PM) and Tomato spotted wilt virus (TSWV). The system is based on a manipulator, which facilitates reaching multiple detection poses. Several detection algorithms are developed based on principal component analysis (PCA) and the coefficient of variation (CV). Tests ascertain the system can successfully detect the plant and reach the detection pose required for PM (along the side of the plant), yet it has difficulties in reaching the TSWV detection pose (above the plant). Increasing manipulator work-volume is expected to solve this issue. For TSWV, PCA-based classification with leaf vein removal, achieved the highest classification accuracy (90%) while the accuracy of the CV methods was also high (85% and 87%). For PM, PCA-based pixel-level classification was high (95.2%) while leaf condition classification accuracy was low (64.3%) since it was determined based on the upper side of the leaf while disease symptoms start on its lower side. Exposure of the lower side of the leaf during detection is expected to improve PM condition detection. © 2016 IEEE.

Note:

Related Files :

Related Content

More details

You may also be interested in