Cargo stacking superelevation and overrun monitoring system based on AI identification technology

Abstract

The invention discloses a cargo stacking superelevation and overrun monitoring system based on an AI identification technology, which belongs to the field of artificial intelligence and comprises an image acquisition module, a calibration reference module, an AI image identification module, a three-dimensional reconstruction module, a real-time comparison module, a structured alarm module and a data return module. The image acquisition module is located above the stacking area, and a multi-baseline camera array is adopted to synchronously acquire images; the calibration module is a grating scale which is arranged in a coplanar manner and provides a sub-pixel-level coordinate reference; the AI identification module extracts a multi-instance mask based on an improved Mask2Former to generate an instance map; the three-dimensional reconstruction module is used for constructing a point cloud by using a PatchMatchStereo algorithm and correcting the depth; the comparison module compares the point cloud with a reference height and identifies an over-limit area; the alarm module outputs level mapping based on a dual-core Gaussian model; the data return module uploads the result through a CAN-FD protocol; the AI recognition and alarm module is provided with an error feedback channel, drives attention correction and forms a closed-loop self-learning mechanism. The method has the beneficial effects that the over-height and over-limit risk in the cargo stacking process is effectively identified and warned.

Images