This phase II work is broken down into three main categories: hardware and integration (~30%), data collection (~25%), and algorithm development and data analysis (~45%). Much of this work will be…

This phase II work is broken down into three main categories: hardware and integration (~30%), data collection (~25%), and algorithm development and data analysis (~45%). Much of this work will be happening in parallel as existing systems are updated and uses as next stage systems are under development. NWB sensors, INC. Implements an iterative scientific / engineering approach on all our projects. This approach follows the steps of 1. Model the problem or design the system, 2. Validate the model or system data by collecting real world data (testing), 3. Refine the model or system if desired performance was not achieved. Throughout the project engineering project management methods will be implemented. These methods include industry standard software tools for: collaborative project management; documentation; software version control, and backup of project documents.development on the improved camera controller will start in fall 2019 and continue into early spring 2020. At this timethe camera controller will be updated to implement improved movement detection and camera control. During development these motion detection and camera control routines will be tested using automobiles in simulated field tests. Work during the early spring will validate these improvements in farm tests. Initial work on the nvidia jetson embedded machine vision platform will also start at this time with the goal of prototype field test during the on-farm experiments in summer 2020. Based on the results of these tests the prototype system will be improved upon during fall and winter 2020 to be ready for field tests in summer 2021.development to improve the data processing toolchain and to implement the end user data application will happen in parallel and will take place throughout the project starting in fall 2019. The first task of this effort will be to implement object tracking in the data processing tool chain. The illumination correction and improved anomaly detection will utilizethe object tracking routines. As new imagery is added during the 2020 and 2021 summer these routines will be improved as necessary. Testing of the methods to communicate in field data will be developed with a prototype ready for summer 2020.critical to the success of this project is the creation of tools to allow for system of learning for new environments. Development on the methods for assisted training using auto-sorted classes can start in fall 2019. This work can use object identification graphs trained on grain crops but being adapted to pulse crops. Validation of these techniques will use comparisons with existing human built graphs for pulse crop object identification. After the motion tracking has been integrated into the tool chain further development on automated new crop and object detection can take place.work to validate data processing and training tool chains will take place throughout the project using both newly collected and existing data. Validation will require close work with our CO-operating growers and landowners to ensure the mapping data is accurate and in formats usable in their operations.