3D Sensor Monitoring Using Projective Shadow Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine monitoring systems using 3D sensors face challenges in accurately determining distances between objects and machines, especially due to projective occlusion and shadowing effects, which can lead to unreliable hazard detection and safety responses in human-robot collaboration scenarios.
Innovation Solution
A method that evaluates 3D image data by considering projective shadows as part of the objects or machine parts to calculate distances, using spherical representations for machine parts and objects, and cone models for shadows, ensuring geometric accuracy and reliability in distance determination, even in the presence of occlusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 3D sensors are used to monitor objects in the machine environment, then the detection ability and safety monitoring capability are improved, but the projective occlusion and shadowing effects cause inaccurate distance determination and unreliable hazard detection
Solution Approach 1:
The monitoring space is divided into multiple detection fields assigned to different 3D sensors. Each sensor is responsible for a specific spatial region, and the evaluation unit coordinates multiple sensors to achieve complete coverage. This segmentation allows each sensor to focus on its designated area, reducing the impact of occlusion and shadowing effects within individual detection fields while maintaining comprehensive monitoring capability.
Solution Approach 2:
An evaluation unit acts as an intermediary between multiple 3D sensors and the safety control system. It receives 3D image data from multiple sensors, processes the information to compensate for occlusion and shadowing effects, determines accurate distances between objects and machine parts, and generates safety-related signals. This intermediary processes the raw sensor data to produce reliable hazard detection results despite the limitations of individual sensors.
2Reliability
If protective fields are configured to ensure safety, then the safety of persons is improved, but the working space for human-robot collaboration is reduced
Solution Approach 1:
The safety monitoring system dynamically adjusts the protective field boundaries based on real-time detection of objects and their movement trajectories. Instead of using fixed safety distances, the system continuously evaluates the actual hazard level by determining precise distances between objects and machine parts, allowing the protective field to expand or contract dynamically. This enables maximum working space utilization while maintaining safety through adaptive, real-time monitoring.
Solution Approach 2:
The system changes the parameters of protective fields from static fixed distances to dynamic values based on real-time 3D object detection and distance calculation. By continuously updating the safety parameters according to the actual positions and movements of objects, the system optimizes the balance between safety protection and working space availability, allowing operators to work closer to the machine when hazard levels are low while automatically enforcing stricter boundaries when hazards are detected.
3Area of stationary object
If multiple 3D sensors are used to cover the entire machine environment, then the detection coverage is improved, but the device complexity and computational overhead increase
Solution Approach 1:
The monitoring task is segmented across multiple 3D sensors, with each sensor assigned to a specific detection field or spatial region. This segmentation allows the system to achieve comprehensive coverage without requiring a single complex sensor to process the entire environment. Each sensor processes a manageable portion of the scene, reducing individual computational loads while collectively providing complete monitoring coverage through coordinated evaluation.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method for monitoring a machine (26) is described in which the positions of machine parts are recorded, 3D image data of an environment (12) of the machine (26) are acquired from an origin position, and the positions of objects (28) are determined from the 3D image data. A distance between the machine (26) and objects (28) in the environment (12) of the machine (26) is determined by calculating the distance between pairs of a machine part position and an object position (28), and the projective shadows (32, 34) of the machine part and/or the object (28) are added for distance determination.