Liczenie zapasów AI za pomocą dronów i widzenia komputerowego

Physical inventory observation has long been one of the most labor-intensive audit procedures. PCAOB AS 2510 and ISA 501 generally require auditors to attend the physical count when inventory is material and attendance is practicable. They observe the count process, make test counts and consider cutoff and condition. Computer vision changes how the count is taken. In warehouses, autonomous drones or vehicle-mounted cameras move along aisles, and vision models decode barcodes, read text labels, and classify each location as occupied or empty. Vendors such as Gather AI and Corvus Robotics sell drone systems for this kind of scanning. Software then reconciles each scan with the warehouse management system and lists matches, mismatches, missing pallets and unexpected ones. Outdoors, photogrammetry or LiDAR builds a 3D surface of piles of coal, grain, aggregate or wood chips and calculates volume against a base surface. Several large audit firms have publicly described trying drones for inventory observation. For the auditor, the key question is whose evidence it is. A drone count run by the company is management's count, so the auditor evaluates the count procedures and the controls over the data, and still performs independent test counts. A tool run by the auditor produces audit evidence whose reliability the firm must establish. Several misconceptions persist. A scanned pallet label identifies the SKU but does not prove how many units are in the shrink-wrapped pallet. Images rarely show obsolescence, damage or whether goods on hand are held on consignment and owned by someone else. A stockpile volume is not a quantity until a density assumption turns it into tons, and that assumption can move the result more than any imaging error.