Learned Obstruction Detection for Machine Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems for machines often produce false-positive alerts due to the inability to distinguish between foreign objects and machine-associated components, leading to unnecessary disruptions and potential damage, especially in dynamic work environments where machine configurations vary.
Innovation Solution
A system utilizing a detection sensor, output device, and controller with a learned obstruction detection process that compares detection signals with component-associated data to differentiate between foreign objects and machine components, reducing false positives by defining an obstruction zone and classifying signals based on machine component occurrences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If object detection systems use sensors to detect objects in the vicinity of a machine, then object presence can be identified, but false-positive alerts are generated when machine components are detected instead of foreign objects
Solution Approach 1:
The system performs preliminary learning during an initial period when the machine is stationary or in a known safe state. During this learning phase, the system detects and stores the radar signatures of machine components (implements, structures, ground engaging devices) before they can cause false-positive alerts during actual operation. This preliminary action enables the system to distinguish between machine components and foreign objects during normal operation.
Solution Approach 2:
The system continuously monitors radar detections and uses feedback to update the learned obstruction detection process. When the machine operates in different configurations or positions, the system receives feedback about which detected objects are actually machine components versus foreign objects, and adjusts the detection algorithm accordingly. This feedback mechanism refines the system's ability to accurately distinguish between the two types of objects over time.
2Reliability
If a general false detection database is used to filter objects, then some false positives are reduced, but the filter cannot accommodate dynamic machine configurations and moving implements
Solution Approach 1:
The system transitions from a static, pre-programmed false detection database to a dynamic, adaptive learned obstruction detection process. The system automatically learns and adapts to the specific machine configuration, implement positions, and operational context in real-time. This dynamic approach allows the system to handle any machine configuration or implement position without requiring manual updates to a database, significantly improving adaptability while maintaining false warning reduction.
Data Source
AI summary
A system for detecting objects in a zone proximate to a machine includes a detection sensor, an output device, and a controller. The controller is configured to define an obstruction zone proximate to the machine and within the zone, receive detection signals from the detection sensor, determine if the detection signals indicate that an object exists within the obstruction zone, and determine if the object is indicative of a machine component, if the detection signals indicate that the object exists within the obstruction zone, based on a learned obstruction detection process. The learned obstruction detection process is configured to determine if the object is indicative of a machine component by comparing the detection signals with component-associated detection data. The controller is configured to provide an alert signal to the output device if the first object is not indicative a machine component.