A multi-source perception fusion system for nuclear environment source positioning and autonomous homing operation

Abstract

The application discloses a multi-source perception fusion system for nuclear environment source item positioning and autonomous trend operation. The system comprises a chassis integrated with a hardware synchronization controller and a sensing array composed of a single-line laser radar, a depth camera and a coded aperture gamma camera. The system realizes microsecond-level time alignment of the sensors through pulse signals. A BEV space-time fusion module maps point clouds and radiation features to a unified bird's eye view coordinate system by using depth probability prediction and view cone projection, and constructs a four-dimensional voxel map containing geometric occupancy and radiation intensity. In view of shot noise, the space-time correlation confidence is calculated by monitoring the dose rate gradient, so that the image is adaptively denoised and enhanced. A dynamic source inversion module integrates a physical consistency verification operator based on a Transformer, uses a cross attention mechanism to verify the geometric and radiation logical relationship, and corrects the positioning artifacts; and combines a ray attenuation model to iteratively solve the three-dimensional coordinates and activity of the radiation source. The application solves the problems of missing perception dimension and positioning ghosting in a nuclear environment, and improves the safety and accuracy of operation.

Images