Aristotle’s “the whole is greater than the sum of its parts” adage rings true for the powerful combination of AI and human insight, as discussed by industry experts – including Trimble’s Tim Lemmon, marketing director for geospatial office software and applications – in a recent article from SPAR3D.
Artificial Intelligence (AI), such as Trimble’s eCognition, continues to make great strides in laser scanning workflows and is a very convenient technology tool; yet, it is not intended as a replacement for humans. The combination of human and artificial intelligence creates the most robust, accurate, efficient and effective use of AI for 3D laser scanning.
As an example, humans help AI computer programs conduct object-based image analysis (OBIA) by teaching the computer to decipher objects, such as trees, signs or poles. Human intervention is needed to “teach” the AI program to recognize variations in these objects. This combination of consistent inputs with human feedback is essentially the way AI and humans work together for the most accurate laser scanning workflows.
Industry experts compare AI to a basic risk-reward calculation, writes author Sam Pfeifle in the article. How accurate is the automation and how much can you rely on it to identify objects in a laser scan? How much human interaction is needed to achieve quality scans? How much risk is incurred by false positives and missed objects?
Lemmon advises that AI programs need to be accurate at least 85-90% of the time, to deliver the needed results of a scan. Without that level of reliability, validation and quality assurance is too time consuming.
“If you’re picking out signs or poles and going through and manually clicking on points to create a CAD line, it’s often faster to do that manually unless you’re getting greater than 85 percent,” said Lemmon.
“You can accept some error if you’re looking at, say, loss of rainforest in the Amazon,” Lemmon added. “But if you’re doing extraction for a civil engineering project, your margin for error is much lower.”