Camera-Based Target Motion Detection With Depth Coordinate Fusion

The patent replaces the LiDAR mechanical scanning system with a camera-based optical system. Instead of using laser beams to scan and collect point cloud data, the system uses image capturing devices to acquire image data, which is then processed to extract target position and motion information. This substitution eliminates the need for mechanical laser scanning while maintaining measurement capabilities through computational methods.

The patent creates a virtual copy of the physical scene through image capture and processing. Instead of directly measuring target position with LiDAR, the system captures images of the scene, processes these images to generate virtual representations (including depth map generation through neural networks), and extracts motion information from these virtual models. This copying approach enables efficient processing while maintaining measurement precision.

The patent replaces the high-energy LiDAR laser emission system with a low-energy camera-based optical capture system. The image capturing device uses ambient light rather than emitting high-frequency laser beams, dramatically reducing energy consumption while obtaining sufficient depth and position information through image processing and neural network-based depth estimation.

The patent enables the system to derive depth information autonomously from captured images without requiring active illumination. The neural network processes the captured image data to generate depth maps, allowing the system to self-service the depth estimation function using passive optical information rather than active laser illumination, thereby reducing energy consumption.

The patent replaces the complex point cloud data processing pipeline with a simplified image-based processing system. Instead of performing target detection and tracking on three-dimensional point cloud data from multiple time points, the system processes two-dimensional image sequences, which are computationally less intensive and require simpler processing algorithms while still enabling accurate motion information extraction.

The patent uses image-based virtual representations to simplify the complexity of physical point cloud processing. By creating and processing virtual image copies of the scene rather than manipulating complex three-dimensional point cloud structures, the system reduces computational complexity and device requirements while maintaining the ability to extract accurate motion information through frame-to-frame comparison.