Machine Vision Robotic Loading With Structured Light Depth Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic pick-and-place systems struggle with chaotic and disordered agricultural commodity arrangements, leading to low throughput and safety risks, as current depth imaging modalities like time-of-flight and stereoscopic imaging are unsuitable for piled commodities due to low depth resolution, obstructed views, and misalignment issues.
Innovation Solution
A machine vision system using a camera, aimable illumination sources, and a controller to capture and process depth maps and background images, combined with machine learning, enables precise detection and manipulation of objects on industrial lines, even in disordered piles, by projecting illumination, capturing images, and determining depth maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current depth imaging modalities (time-of-flight or stereoscopic imaging) are used, then the system can provide depth information, but the depth resolution is low and the field-of-view is obstructed
Solution Approach 1:
The patent transitions from conventional 2D image capture to structured light scanning that captures 3D depth information. By projecting structured light patterns and analyzing their deformation on objects, the system achieves high-resolution depth mapping without obstruction issues, as the structured light can wrap around and illuminate hidden surfaces that conventional cameras cannot capture.
2Adaptability or versatility
If manual labor is used for disordered piled commodities, then the system can handle random arrangements, but the throughput is low and labor safety is compromised
Solution Approach 1:
The system uses machine learning models that automatically learn to identify and locate objects in disordered arrangements without requiring pre-programmed knowledge of specific configurations. The model processes structured light scan data to autonomously determine object positions, orientations, and types, enabling the robotic system to adapt to any random arrangement while maintaining high throughput and eliminating manual labor requirements.
3Manufacturing precision
If prior knowledge of localization and orientation is required, then the system can achieve precise manipulation, but the system cannot handle random or chaotic arrangements
Solution Approach 1:
The system performs preliminary structured light scanning and depth map generation before robotic manipulation. The machine learning model processes this pre-captured data to determine precise localization and orientation of all objects in the scene, allowing the robotic system to plan and execute manipulation tasks with high precision while handling completely random arrangements, as all necessary spatial information is obtained in advance through non-contact scanning.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides sub-millimeter resolution and reduced height limitations, allowing for accurate reconstruction of textured and untextured products, and enables robotic loading without conveyor movement, improving throughput and safety.
Implementation Method 1
control the camera to capture a scan image comprising the projected illumination
Data Source
AI summary
Systems, methods, and media for machine vision-guided robotic loading are provided. Such systems and methods can allow for more accurate object detection so as to allow for robotic picking or loading of an object from a pile or group of objects. In some embodiments, a camera is controlled to capture a background image of the object or group of objects. One or more illumination sources (such as, e.g., laser illuminators) are controlled to project immunization toward the object or group of objects within the field-of-view of the camera. The camera may be controlled to capture a scan image comprising the projected illumination. Based on the scan image, a depth map can be determined, which is provided to a machine learning model along with the background image. The output of the machine learning model can then provide information associated with one or more objects, such as location information, orientation information, and the like.