Anomaly Picking Point Detection for Robotic Material Sorting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sorting methods fail to effectively remove irregularly shaped anomalies from material collections, leading to decreased product value and equipment damage.
Innovation Solution
A method using a processor to identify anomalies, generate bounding boxes, and determine picking points on anomalies for accurate removal by robotic means, employing image analysis, segmentation, and machine learning to ensure picking points are on the anomaly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional sorting methods are used, then the sorting process is simple, but irregularly shaped anomalies cannot be effectively removed
Solution Approach 1:
The patent replaces traditional mechanical sorting methods with a computer vision-based system that uses image processing, machine learning, and automated robotic manipulation. The system captures images of material, identifies anomalies through AI algorithms, calculates precise picking points, and uses robotic actuators to remove anomalies, thereby achieving high reliability in removing irregularly shaped anomalies without relying on simple mechanical sorting
Solution Approach 2:
The system employs machine learning models that are trained to automatically identify and classify anomalies without human intervention. The anomaly detection model autonomously analyzes images, determines anomaly characteristics, and the picking point calculation algorithm automatically computes optimal removal points, enabling the system to serve itself in identifying and removing anomalies consistently
2Manufacturing precision
If picking points are not accurately located on anomalies, then the system operates simply, but anomalies are not completely removed leading to product degradation
Solution Approach 1:
The system performs preliminary image capture and analysis before the actual anomaly removal action. It captures images of the material, processes them through machine learning models to identify anomalies, calculates optimal picking points on each anomaly, and only then executes the removal action. This preliminary analysis ensures high picking point accuracy by thoroughly understanding anomaly characteristics before intervention
Solution Approach 2:
The patent replaces simple mechanical picking with a sophisticated computer vision and robotic control system. The system uses image processing to locate anomalies, applies machine learning to understand anomaly characteristics, calculates precise picking points considering anomaly shape and position, and controls robotic actuators to execute removal with high precision, thereby achieving manufacturing precision in picking point accuracy
3Reliability
If irregularly shaped anomalies are not completely removed, then the sorting process remains fast, but product value decreases and equipment damage occurs
Solution Approach 1:
The patent replaces traditional mechanical sorting with an automated vision-guided robotic system that captures images, identifies anomalies through machine learning, calculates precise picking points, and executes removal actions. This automated system ensures complete anomaly removal by verifying picking points are accurately located on anomalies, preventing product degradation and equipment damage while maintaining sorting throughput through automation
Solution Approach 2:
The system implements a feedback loop where images of material are captured, anomalies are identified and classified, picking points are calculated, removal actions are executed, and the results can be verified through continued image monitoring. This feedback mechanism ensures complete anomaly removal by continuously monitoring the material stream and verifying that removal actions were successful, while maintaining productivity through automated real-time operation
Data Source
AI summary
A processor may receive an image of the collection of material having a plurality of objects. The processor may identify the anomaly from the plurality of objects. The processor may generate a bounding box for the anomaly. The processor may generate one or more picking points on the anomaly. The one or more picking points may be configured on at least one balance points of the anomaly. The processor may remove the anomaly from the collection of material via the one or more picking points.