This paper describes an approach to measuring sugarcane yield on a sugarcane chopper harvester using a volumetric flow-based measurement of the harvested product on the elevator, and then converting to mass flow using a density that was found by calibration. Initial proof of concept testing on a stationary setup with a John Deere 3520 sugarcane chopper harvester was carried out using bamboo as a surrogate material due to concerns of spoilage with sugarcane billets. Results showed a strong correlation between the detected volume flow and mass flow, with an R-squared of 97%, and a 4.6% coefficient of variation of measured density values. These positive results led to further field testing in Brazil and Louisiana during 2013, and Texas during the 2014 harvest season. Data was gathered in a wide range of field and operating conditions to identify the full predictive capabilities of the system, as well as identify limitations. Analysis of the data collected from field experiments revealed an interesting and useful relationship between volume flow and bulk density, in which bulk density of material on the elevator decreases along a curve that has the same characteristic shape as √x/x. The decrease in density is likely due to less cleaning/removal of trash at higher volume flows through the machine, as well as changes in the way the billets pack in the slats (more loosely packed at larger volume flows). This trend was used by applying a square-root transformation on the measured volume values, which in turn linearized the relationship between volume flow and mass flow such that a simple linear calibration factor (density) could be used to convert measured volume flows to predicted mass flows. After applying the transformation, the average coefficient of variation of measured density values with a given field was about 5.6% for green cane, with indications that extreme variations in machine operating settings, such as fan speed, that were induced during testing caused the average coefficient of variation to be slightly higher than would be expected during typical harvesting. Burnt cane also benefited from transformation of the volume values, albeit not quite as much, and had an average coefficient of variation of roughly 6.1% in density values. However, at least some of the variation can be attributed to the fact that burnt cane can tend to have a large variation in trash content (pockets of trashy burnt cane mixed in with clean burnt, and vice versa), which in turn causes large fluctuations in measured density values. Yield plots of several fields that compare actual to predicted show very little difference, as long as an appropriate calibration scheme is enacted to ensure a reasonable estimation of the density value for a given field. The research shows great potential for commercialization as a yield monitor for the sugarcane industry, which has not yet seen one of suitable accuracy.