April 27, 2018

It’s Easy, But Is It Accurate?

Author: Stockpile Reports

Some have a hard time believing that walking around a pile with an iPhone or flying a drone capturing pictures could possibly result in reliable information about the volume of a pile without touching it.

Stockpile Reports ACCURATELY calculates volume.

What you really want to know is, “is my inventory right?”  Measurements rely on consistency, density, pile management, site condition, but system accuracy is usually the smallest variable in your assessment of accuracy.  So we’d like to broaden the discussion to include considerations affecting repeat-ability, consistency (getting the same measurement each time).

You probably know what it’s like to rely on different people to mark piles and get different measurements each time, so let’s review why, with our system, you WILL know your numbers when it comes to volume.

How do I know the measurement is true?

Simple geometric equations on a broad scale result in high confidence.

Volumetrics is a simple matter of geometry and point calculations on the pile. The more points, the better. In a typical Stockpile Reports measurement, there are hundreds of thousands of points captured. This not only senses the overall shape of the pile and surrounding terrain but also models subtle variances in the surface and contours.

In this case, accuracy is dependent on the resolution and quality of the images which is why we recommend specific high-resolution cameras and capture techniques.

How are you auto-detecting the base of a pile?

We use computer vision to analyze and detect the base of the stockpile. We have taught our system how to recognize a pile as distinct from the ground which lends to consistency from pile to pile. Many other processes require a person to manually mark the pile and the results can vary from person-to-person.  Our system automatically picks tens of thousands of points to detect the base. No human variance is in play, so the assessment is not subjective. The same algorithm and process are applied time and again to repeatedly and consistently detect the pile’s base. When uncertain, the system will mark the pile in red as an alert. (such as when there is vegetation, standing water, or obstructions).  Compare this to hand-marking piles where a different person may have different techniques for marking piles, giving you variances on your measurements that are not coming from volume changes but measurement approaches.

How does the system calculate tonnage?

Our system utilizes a conversion rate that you enter to calculate weight from the volume.  Most measurements range from 1.2-1.3 and generally go from .5 to over 2 tons per cubic yard.  We recommend ASTM Standardized Volume Testing.  Here is an example for course material calculations.

LiDAR comparison

How does 3D vision technology like Stockpile Reports compare?  Stockpile Reports is within 1.6% of LiDAR which utilizes millions of laser points but requires a trained surveyor, ground control points, and manual pile markings.  Our point density with images is similar but computer vision interprets the data as part of its intelligence.

As for the number of points required to model a pile, tens of thousands of points are typically sufficient. To get more points than needed is negligible in terms of accuracy.

On the other end of the spectrum, those familiar with Range Finders know that they typically rely on under 50 points per pile which can miss key attributes of a pile shape in the calculation.

How does it compare to GPS/GIS solutions?

Stockpile Reports isn’t for traditional surveying. If you’re trying to combine inventory measurements with GPS location data, you may wish to utilize a different system than ours —one that is suited for mine planning, GIS-engineering-focused solutions, blasting positions, or when you wish to merge GPS data with the aerial images.

GPS points themselves are accurate as they relate to a position on the earth. However, a typical inventory measurement based on GPS tools will contain at most a few hundred points to represent the pile’s surface. This causes gaps in the final model of the pile, and key parts of the pile’s shape and contours can be missed within GPS-based approaches.

Compared to GIS/GPS solutions, we bring the same level of confidence in the volume of a pile and we’ve automated many of the steps necessary in a survey including pile detection, marking, and report generation. Our system does not require ground control points which can take considerable time (and expertise) to set.

Compare the time savings with GPS/GIS drone solutions vs. Stockpile Reports where we fly for you.  All that’s left is to schedule, review and approve.

What else can affect accuracy?

The condition of the site and piles can impact the accuracy of your measurement. See Preparing Your Site for Inventory guide to make sure your site conditions are set to give you the best measurement outcomes.

Several factors are known to influence measurements more than the technology itself including:

• Density: A key number used to calculate the weight/tonnage. Conversion factors are, by far, the #1 culprit for inaccuracies concerning the final tonnage of a pile.  See the Standard Test Method for calculating densities of course materials
• Consistency of the pile markings:  Is there a consistent methodology?  Is it marked by a human? Did they estimate the toe?  With our system, the toe is automatically detected using consistent algorithms.
• Environmental factors:  Standing water, obstructions, and vegetation can block your visual view of the pile and your measurements. With our system, you’ll know when there is a risk.
• Layout of the pile:  If you use traditional measurements, it’s difficult to collect enough sample points to accurately calculate variations in the top of the pile.  Even with Stockpile Reports, a pile with a jagged top must be visible to the camera. If your image collection does not capture the top, your accuracy could be affected by missing information.

What have other customers found when they compare Stockpile Reports?

Our system has been tested by many of our customers and government agencies with positive results.

To truly understand stockpile measurement accuracy we’ve conducted multiple 2-day workshops where we bring a terrestrial laser scanner and perform various measurements with the Stockpile Reports method and existing methods and perform measurements repeatedly with different people to assess precision. We can also build piles to run across the scales.

We ran this process recently under the supervision of KPMG to learn that consistent methods for determining base planes and dealing with obstructions are the largest drivers of accuracy and precision.

Often, when we review contours or point clouds to identify differences in methods, it’s difficult to determine which one is “right”.  We have an expert from the site review the “unverified measurements” to confirm that the system made appropriate judgment calls. In this case, these adjustments are “learned” and used in the future to improve measurements over time.

We’ve had dozens of internal and external tests, but the most widely-published was the AASHTO Results from Award – TX DOT utilized 30 measurements comparing iPhone to LiDAR – comparable volume measures were within 1.6% of LiDAR.