TRADITIONAL CROP BREEDING RELIES ON TIME-CONSUMING AND INEFFICIENT MANUAL FIELD OBSERVATIONS OR SAMPLINGS. UNMANNED AERIAL VEHICLE (UAV)-BASED HIGH-THROUGHPUT PHENOTYPING PROMISES TO ALLEVIATE THE PHENOTYPING BOTTLENECK AND PROVIDE THE CAPABILITY FOR REPEATED NON-DESTRUCTIVE MEASUREMENT OF THOUSANDS OF GENOTYPES GROWN IN FIELD EXPERIMENT PLOTS IN CROP SELECTION CYCLES. MAJOR CHALLENGES REMAIN TO ENABLE MORE SOPHISTICATED ANALYSES AND DEVELOPMENT OF INTEGRATED DECISION-MAKING SYSTEMS THAT CAN GREATLY ACCELERATE CROP IMPROVEMENT AND PHENOTYPING. WE PROPOSE TO DEVELOP A DIGITAL RICE SELECTION SYSTEM THAT INTEGRATES UAV IMAGING, MACHINE LEARNING, AND MULTI-TRAIT DECISION-MAKING. OBJECTIVES ARE: 1) QUANTIFY KEY PHENOLOGICAL, MORPHOLOGICAL AND ARCHITECTURAL TRAITS THAT CAPTURE RICE GROWTH AND DEVELOPMENT; 2) ACQUIRE MULTI-VIEWPOINT UAV IMAGES OF RICE GENOTYPES DURING CRITICAL GROWTH STAGES; 3) DEVELOP ADVANCED IMAGE ANALYSIS ALGORITHMS TO DERIVE KEY PHENOLOGICAL, MORPHOLOGICAL AND ARCHITECTURAL TRAITS FOR CRITICAL RICE GROWTH STAGES, AND 4) DEVELOP A DIGITAL RICE SELECTION SYSTEM THAT SCREENS FOR BEST-PERFORMING GENOTYPES THROUGH DATA INTEGRATION AND MULTI-TRAIT DECISION MAKING.GROUND TRUTH DATA FOR KEY TRAITS DURING CRITICAL RICE GROWTH STAGES WILL BE COLLECTED ALONG WITH HIGH-RESOLUTION RGB/MULTISPECTRAL UAV IMAGES FROM MULTIPLE CAMERA ANGLES. WE WILL DEVELOP (1) ADVANCED MACHINE LEARNING ALGORITHMS TO EXTRACT KEY TRAITS, (2) MULTI-TRAIT-BASED MACHINE LEARNING MODELS TO ESTIMATE FINAL ABOVEGROUND BIOMASS AND GRAIN YIELD, AND (3) A MULTI-CRITERIA DECISION-MAKING SYSTEM TO SELECT BEST-PERFORMING RICE GENOTYPES. THE PROPOSED PROJECT REPRESENTS A MAJOR EFFORT IN DELIVERING AN INTEGRATED UAV IMAGERY-BASED DECISION-MAKING SYSTEM TO RICE BREEDERS AND RESEARCHERS AND WILL BE AN INDISPENSABLE TOOL TO GREATLY IMPROVE RICE BREEDING AND PHENOTYPING EFFICIENCY.