Container Crane Spreader 3D Point Cloud Target Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current container crane control systems face challenges in accurately identifying landing targets, especially in conditions with limited visibility or when the target is situated higher than surrounding surfaces, which can lead to inefficiencies and potential inaccuracies in container placement.
Innovation Solution
The method involves obtaining two-dimensional images from multiple cameras on the spreader, performing feature extraction using techniques like SIFT, and generating a three-dimensional point cloud to identify key features such as corners or twistlocks of the landing target, allowing for precise control of the container's movement based on the point cloud data and orientation detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional single-camera or simple image processing is used, then the system complexity is low, but the landing target identification accuracy deteriorates in adverse weather or high-position targets
Solution Approach 1:
The patent transitions from two-dimensional image processing to three-dimensional point cloud representation. Multiple cameras capture 2D images that are then processed to generate a 3D point cloud model of the landing target, enabling accurate identification of key features (corners, twistlocks, gooseneck) even when the target is positioned higher than surrounding surfaces or in adverse weather conditions. This dimensional transformation provides depth information and spatial context that 2D images alone cannot provide.
Solution Approach 2:
The patent divides the landing target identification task into multiple segments: capturing 2D images from multiple cameras, extracting features from each image, matching features across images to build a 3D point cloud, and identifying key features based on spatial relationships. This segmentation allows the system to handle complex targets by processing them in manageable parts and combining the results.
2Measurement precision
If feature extraction algorithms like SIFT are applied, then the key feature identification improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary feature extraction using algorithms like SIFT on each 2D image before generating the 3D point cloud. By pre-identifying potential key features (corners, edges, distinctive points) in each individual image, the system can then efficiently match these features across multiple images to construct the 3D model, reducing the overall processing time compared to analyzing the complete 3D structure from scratch.
Solution Approach 2:
The patent creates a simplified representation (point cloud) that copies only the essential spatial information from the complex 2D images. This abstraction retains the key geometric features and spatial relationships while discarding unnecessary visual details, enabling faster processing and more efficient feature matching compared to working with the original high-resolution images throughout the entire pipeline.
3Reliability
If multiple cameras are used to capture images from different positions, then the three-dimensional point cloud generation improves, but the device complexity increases
Solution Approach 1:
The patent merges the data from multiple cameras by capturing images from different positions and angles, then combining these 2D images through feature matching to generate a unified 3D point cloud representation. This merging process integrates information from all cameras to create a comprehensive and reliable model of the landing target, improving identification reliability by providing multiple viewing perspectives and redundant information.
Data Source
AI summary
A method for loading a container on a landing target on a land vehicle using a container crane including a trolley and a spreader for holding and lifting the container. The method is performed in a container crane control system and includes the steps of: obtaining two-dimensional images of the landing target from a first pair of cameras arranged on the spreader; performing feature extraction based on the two-dimensional images to identify key features of the landing target; generating a point cloud based on the feature extraction, wherein each point in the point cloud contains coordinates in three dimensions; and controlling movement of the container to the landing target based on the point cloud and the identified key features of the landing target.