Dataset for precision yield estimation and mapping in manual strawberry harvesting

This dataset contains harvest data collected during manual strawberry harvesting using instrumented picking carts (iCarritos) in two California fields over a harvest season. The data includes geo-tagged harvest mass and cart motion information recorded by load cells, a GPS receiver, and Inertial Measurement Units (IMU). The dataset enables data-driven yield estimation and mapping for manually harvested strawberries. Furthermore, the dataset could be used to optimize robotic strawberry harvest machines or harvest assist systems, such as crop transportation systems. This dataset is valuable for researchers and practitioners in precision agriculture, offering a foundation for developing scalable automation and precision agriculture technology in specialty crop harvesting. 

Methods
Data was collected from iCarritos developed by modifying a conventional wireframe structure with a wheelbarrow system, integrated with sensors, mounting systems, and control hardware. Two load cells were installed at the front and back of each iCarrito, and the harvest mass was obtained by averaging the readings from these load cells. A significant design choice was to ensure that the iCarritos were low-cost and practical. To achieve this, the carts utilized an affordable Piksi Multi GNSS unit (Swift Navigation Inc., USA) with an integrated IMU. The GNSS unit operated based on SBAS localization and has a rated horizontal Circular Error Probable (CEP) accuracy of 0.75 meters. This unit was used to measure location (latitude and longitude) as well as acceleration data along the x, y, and z axes (ax, ay, az). 

A Raspberry Pi Zero W (Raspberry Pi Foundation, UK) microcomputer served as the central processing unit, with an SD card used to run the cart software and store data during harvest operations. The data was stored at the rate of 10 Hz. The control system consisted one main power toggle switch connected directly to the battery, a control switch for selecting operation modes, and two status LEDs for system feedback. 

Before taking the iCarritos into the field, each cart's load cells were calibrated. Following each harvest, the collected data was reviewed and the load cells were recalibrated if necessary. Commercial strawberry pickers, who were experienced in manual harvesting but not specifically trained on the iCarritos, utilized these instrumented carts during normal harvesting operations. The pickers were compensated for their time using the iCarritos in the trial field. They were instructed to use the iCarritos as they would their regular carts, with minimal changes to their normal harvesting routine. 

Studies were conducted in two major strawberry-producing regions in California during the 2024 strawberry growing season. The selected fields were commercial strawberry fields in Santa Maria, growing the Fronteras variety, and in Salinas, CA, growing the Cabrillo variety. The width of the raised beds in Santa Maria were 110 cm wide while those in Salinas were 75 cm wide. The raised bed centers were also pre-mapped using RTK-GNSS, providing centimeter-level accuracy under field conditions.