All About Precision: Agerpoint Achieves 99% Accuracy In Tree Counting And 95% Accuracy In Canopy Metrics

While we pride ourselves in collecting agricultural data in a timely manner (and with less manpower) than other data collection methods, our main focus is on the accuracy of the data we collect and provide to our clients. Precision is crucial to us here at Agerpoint.

We work with clients, across multiple crop types, who use the data we collect and process for a variety of business, technical, and scientific needs. This requires our data to be precise so it can provide actionable information for our clients.

How accurate can we get?
Pretty accurate – we are able to achieve a 99% accuracy in tree counting, and can achieve a 95% accuracy in canopy metrics, such as canopy diameter, volume, density, and tree height. And if we have clear visibility of trunks, we can also achieve a 95% accuracy of truck diameter.

How do we get such precise and accurate data?
Through a combination of algorithms, QAQC using custom built tools, machine learning and classification methodologies, we are able to get the accuracy as described above. And we continue to validate and calibrate our methodologies on a regular basis for the most precise results.

QAQC Edit Process

The end-to-end Agerpoint data process includes an edit function, the QAQC Edit Process, executed by our trained Geospatial engineers and utilizing proprietary and custom Agerpoint tools. These tools allow our team to manage “outlier plants” identified by the algorithms that process the raw data we collect in the field.

Our QAQC Analyst makes visual edits using our custom tools on these rare outliers, further ensuring total data accuracy. At least ~5% of all scanned trees are sampled and visually analyzed to ensure data precision.

Agerpoint identifies outliers in the data during the QAQC Edit Process.

Field Ground Truthing

As part of Agerpoint’s standard field collection processes, we have developed reference data sets based on crop types from statistically relevant samples of trees and their associated parameters (height, trunk diameter, canopy diameter, etc.). These samples are physically measured and recorded in the field. This data is makes up the ground-truth data set used to benchmark the macro results of Agerpoint’s end-to-end data process.

We compare our collected measurements against the ground-truth data sets. The collected measurement is considered correct if it falls within a threshold of distance of the ground-truth measure (typically around 10%). Comparing our measured data to these data sets it where we are able to derive our accuracy metrics.

Field ground truthing techniques of manually collecting data for comparison against our standard field collection processes to ensure accuracy.

Algorithm Benchmarking

With over 1.5 trillion data points collected across 24 different crop types, we have a large store of historical data. This allows us to compare data sets of like crop types to ensure that what is collected is consistent across the software development process and past data collections.

With our plant matching algorithms, we are able to scan the same plants again and again to further refine the accuracy for an individual client down to the individual plant level. Over time trends or issues may be easier to spot with our data collection than with previously used methods, like manual sampling.

Our methods together are what lead to incredibly accurate data for our clients. Are you interested in learning how you could get 99% data accuracy, with less manpower?