With the development of robotics technology today, AMR (autonomous mobile robot) has become the core driving force in logistics, manufacturing, medical and other fields. These robots are able to navigate autonomously, avoid obstacles and perform tasks, greatly improving efficiency and flexibility. It is their embedded cameras that imbue AMRs with this intelligence. The camera is the "eye" of the robot, and its selection and performance directly determine the reliability and application boundaries of the AMR.
As a consultant specializing in camera modules, this article will provide an in-depth analysis of the two main types of cameras used in AMRs: 2D vision and 3D vision. We will detail key technical considerations when selecting cameras for AMRs, including shutter type, interface options, and 3D vision technology, providing a professional selection guide for embedded vision engineers.
Two Broad Types of Cameras Used in AMRs
In the AMR field, embedded cameras are primarily divided into two categories: 2D vision cameras and 3D vision cameras. Although both are used for environmental perception, their functions and application scenarios are fundamentally different.
1. 2D Vision Cameras for AMRs
These cameras are the common cameras we see every day, primarily capturing two-dimensional image information. They are one of the most basic and important perception sensors for AMRs.
Typical applications for 2D vision cameras include visual SLAM (for autonomous navigation and localization), QR code or barcode recognition, and simple object identification and tracking. They are low-cost and simple to process, making them the core of many AMR navigation systems.
2. 3D Vision Cameras for AMRs
These cameras not only capture images but also acquire depth information about the scene to build a three-dimensional model. This enables robots to perceive the size, shape, and distance of objects.
Typical applications for 3D vision cameras include precise obstacle avoidance in complex environments, precise positioning of pallets or shelves, and grasping tasks for picking robots. 3D vision provides robots with richer environmental data, enabling them to perform more advanced tasks.
Key Factors to Consider When Choosing a 2D Vision Camera
When selecting a 2D vision camera for an AMR, engineers must weigh several key factors. This not only affects image quality but also directly impacts the robot's performance and reliability.
1. Shutter Type: Rolling Shutter vs. Global Shutter Robot Vision
Shutter type is the cornerstone of robot vision. A rolling shutter scans the image line by line, resulting in a "jello effect" or skewed image when the robot moves at high speeds. This is a critical issue for AMRs, which require precise navigation and object recognition.
In contrast, a global shutter captures the entire image simultaneously, ensuring distortion-free images even at high speeds or when capturing moving objects. For AMRs that need to detect moving obstacles or operate in dynamic environments, a global shutter is a more reliable option, although it generally comes at a higher cost.
2. Sensor Resolution and Frame Rate
Higher resolution provides greater detail, which is crucial for QR code recognition, text reading, or detecting small obstacles. However, increased resolution often reduces frame rate and increases processor load. Engineers need to strike a balance between resolution and frame rate to ensure the robot can process image data in real time and respond quickly.
3. Lens Field of View (FOV) and Distortion
The field of view (FOV) of a 2D vision camera determines the range of the robot's environment. A wide FOV is crucial for robot navigation and mapping. However, wide-angle lenses often introduce image distortion, which requires correction through software algorithms; otherwise, navigation accuracy may be affected.
4. Interface Options: Camera Interface Options (USB, MIPI CSI, GMSL2, GigE) for AMRs
The choice of camera interface directly impacts data transfer rate, cable length, and system complexity.
The MIPI CSI interface offers high bandwidth and low power consumption, making it ideal for lightweight embedded cameras for AMRs. However, its cable length is limited.
The USB interface is versatile and easy to use, but it may consume more processor resources and has bandwidth limitations when multiple cameras are used simultaneously.
The GigE (Gigabit Ethernet) interface supports long-distance transmission and is very stable, but it consumes relatively high power and may require an additional network card.
The GMSL2 (Gigabit Multimedia Serial Link) interface is an automotive industry standard that supports long cables and multi-camera transmission, making it an ideal choice for complex AMR systems. However, it comes at a higher cost.
Key factors to consider when choosing a 3D vision camera
In addition to the factors mentioned above for 2D cameras, when selecting a 3D vision camera for an AMR, it's important to focus on the following technical features.
1. 3D Technology Types: Stereo Vision, Time of Flight, and Structured Light
Stereo vision uses two cameras to simulate the human eye, obtaining depth information through parallax calculations. Its drawbacks are that it requires rich textures to function and is computationally intensive. Its selling point is that it is passive and unaffected by ambient light, making it suitable for outdoor applications.
Time of Flight (ToF) calculates distance by measuring the round-trip time of a light pulse. Its selling points are high real-time performance and minimal computational effort. Its drawbacks are that it typically has low resolution and is susceptible to interference in strong outdoor light.
Structured light projects a specific pattern onto a scene and then calculates depth by analyzing the distortion of the pattern. Its selling point is high accuracy. Its drawbacks are significant susceptibility to ambient light and a limited operating range.
2. Depth Accuracy and Effective Range
The depth accuracy and effective range of a 3D vision camera are its most important performance indicators. Picking robots require extremely high depth accuracy to identify and grasp objects, while navigation and obstacle avoidance require a longer effective range. Engineers need to find the optimal balance between accuracy and range to meet the specific needs of choosing a camera for warehouse AMRs.
3. Processor Requirements and Power Consumption
3D vision typically requires significantly more raw data processing than 2D vision. Both binocular disparity calculation and point cloud data processing require a powerful processor. This presents a significant pain point for battery-powered AMRs. Engineers need to consider whether the camera module has a built-in 3D processor and whether its software development kit (SDK) is efficient to ensure the robot's battery life and performance.
Summary
Choosing an embedded camera for an AMR is a complex technical decision that requires a deep understanding of the respective strengths and limitations of 2D and 3D vision. From choosing between a rolling shutter and a global shutter to balancing camera interfaces, every step is crucial. Choosing the right camera is fundamental to reliable robot operation and crucial to project success.
Muchvision helps with AMR selection
Struggling with choosing the right AMR camera for your project? Contact our expert team today and we'll provide you with professional camera modules and embedded vision solutions to help you build a high-performance AMR!
