Insect Monitoring and Early Detection system in Rice During Storage, 2018

 

Zhongli Pan, adjunct professor, Dept. of Biological and Agricultural Engineering, UC Davis

The goal of this project is to develop a new insect monitoring and early detection system for rice storage facilities. Engineers created and tested a “real time” system consisting of insect traps, USB cameras, LEDs, a tiny computer called a Raspberry Pi, a server and a user interface.

Three traps were designed and built from a stainless steel perforated sheet. Each trap was equipped with a USB camera and two LED lights. The lights were placed around the camera and then adjusted for brightness, contrast, and gamma to maximize image quality. Different background colors (white, green, and yellow) in the collecting chamber also were tested. Although there were no significant differences between colors, white was selected for its positive visual effects.

The USB cameras and the LED lights were installed inside the insect traps to provide remote access to images of captured insects. The system was programmed to take images periodically. However, additional images could also be requested through a “take picture” button on the mobile application or website.

When the images were taken, the Raspberry Pi processes them to determine the number of insects. Data and images were sent to a cloud storage service. The server processes the images and counts the insects captured per trap with an insect-counting algorithm. The website and mobile application were developed for the user to be able to view an image gallery for each trap, view a graph of the insects collected over time, and also to request on-demand images.

This new system was evaluated through experiments conducted in the lab and at a commercial rice storage facility in Knights Landing. Lab results and commercial storage tests were consistent and showed that the new system was effective and accurate for monitoring and early detection of insect activity in stored rice. The system detected first insect emergence within 15 to 30 minutes in lab testing at different infestation concentrations.

During the storage test, the system was able to detect emergence within 34 minutes. After four days, the system detected insects in three traps with counting accuracy rates of 75%, 82% and 100%. These results suggest that the new system could be used to detect early stage insect activity in stored rice with reasonable effectiveness and accuracy.

Further research is needed to upgrade the system and improve its handling, customization, effectiveness, accuracy, and energy use before commercialization is likely. The design of traps could be improved with the addition of a wireless system, sensors for temperature and humidity, and an independent power source such as rechargeable batteries. The accuracy of the algorithm also needs some fine-tuning.

These refinements should eventually lead to a new tool for real time monitoring and early detection of insect activity in rice storage facilities with high accuracy, reliability and safety, as well as low cost and labor savings.