Real-time crop disease detection by using UAVs and AI

Crop disease detection can provide information for making disease control decisions. Currently practice of crop disease detection is done by experienced humans which relies on visual assessment and use lab-based specialized equipment to do quantitative analysis. However, both two detection methods are time consuming, and only cover limited areas and representative samples, which lead to inaccurate, subjective, and potential destructive decisions. With the development of Unmanned Aerial Vehicles (UAVs), and in the context of precision agriculture, crop disease detection by using UAVs can cover larger areas. The sensors on UAVs, and analysis by Artificial Intelligent (AI) methods enable the detection results objective and accurate. Although studies on UAV crop disease detection increased in past decade, there are yet unsolved issues, such as real-time detection and the models generalization. Real-time crop disease detection onboard UAV not only can save the labor resources, but also can improve the response speed of decision-making to provide possibility for treatment during detecting. Thus, this PhD project aims to build a real-time detection system for crop disease based on edge-computing technology, and to detect disease from field level to plant level by using lightweight AI models on UAVs with multispectral and visible sensors. Based on these objectives, real-time detection system design and evaluation, development and generalization of onboard lightweight real-time detection model, and implementation and application of detection model on UAVs are the main research question.