Work Machine Object Detection With Movable-Implement Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems in work machines fail to differentiate between work implements and external objects, leading to undesirable alerts and loss of usable data when work implements are present in the perception field.
Innovation Solution
A method and system that generates a multidimensional manifold based on depth data during a calibration stage to distinguish work implements from external objects, allowing for intervention feedback and alert generation only when necessary, thereby minimizing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object detection systems are used in work machines, then the system can detect objects in the perception field, but it cannot differentiate between work implements and external objects, leading to false alerts
Solution Approach 1:
The system segments the perception field into multiple depth layers using depth data from sensors. By dividing the field of view into different depth ranges, the system can distinguish objects at different distances from the work machine, allowing it to differentiate between work implements (closer objects) and external obstacles (更远 objects), thereby eliminating false alerts while maintaining detection accuracy
Solution Approach 2:
The patent introduces depth information as an additional dimension to the traditional 2D image data from perception sensors. By incorporating depth data that indicates the distance of objects from the work machine, the system creates a three-dimensional understanding of the environment, enabling it to distinguish between work implements and external objects based on their spatial position, thus improving both detection accuracy and alert reliability
2Quantity of substance
If work implements are included in the perception field, then the system can detect all objects in the field of view, but it loses usable data due to false positive detections
Solution Approach 1:
The system extracts and removes work implements from the set of detected objects by using depth data to identify and filter out objects that correspond to the work implement's known position and dimensions. This extraction process allows the system to maintain comprehensive detection coverage of the perception field while eliminating false positive detections, thereby preserving usable information for external object detection
Solution Approach 2:
The patent uses depth data as an intermediary parameter to distinguish between work implements and external objects. By introducing this intermediate layer of information that indicates spatial distance, the system can mediate between the complete detection field and the useful detection data, allowing it to retain coverage of all objects while filtering out false positives through the intermediary depth information
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A work method and method provide intervention feedback based on spatially proximate objects in a work area. The work machine includes perception sensors and a work implement. During a calibration stage, the implement is moved to various positions across an available range of movement, and depth data are provided for each position and for each of various perception field portions with respect to objects within a perception field. For each position, a multidimensional manifold is created and stored comprising depth data associated with the work implement for the perception field portions including the work implement. During a machine operation stage, further perception inputs are used to determine current depth data for the perception field portions, and feedback signals are conditionally generated corresponding to an intervention event state which is determined while ignoring any objects identified as corresponding to the multidimensional manifold corresponding to a current position of the work implement.