Industrial 3D vision: technologies, robot guidance, and bin picking

Industrial 3D vision adds depth to measure, inspect, or guide a robot when 2D vision is no longer sufficient, particularly for bulky parts, three-dimensional features, de-bunkering, and robot guidance. This guide explains how it works, its four main technology categories, and the applications where it is making a tangible difference in production: robot guidance using 3D vision, de-bunkering, and dimensional quality control.

What is industrial 3D vision?

Industrial 3D vision encompasses technologies that measure the depth of a scene, in addition to its surface, so that an automated system can precisely control, measure, or guide a robot.

A conventional 2D camera captures a flat image. A 3D camera adds a third piece of data: the distance of each point from the lens. The result is called a point cloud: a three-dimensional map of the observed scene.

This map makes it possible to reconstruct an object’s exact position, orientation, and volume. The sensors used vary: stereoscopic, time-of-flight, structured light, or laser triangulation. Increasingly, they are paired with deep learning models to recognize complex shapes, even on parts that have never been seen before.

In production, three applications dominate: dimensional quality control, precision measurement, and robot guidance, particularly for picking up bulk parts. 2D vision cannot handle this last task on its own.

This should not be confused with human 3D vision, which describes the perception of depth by the brain and the retina. Here, the focus is purely industrial: cameras and algorithms applied to a production line, not biology.

CEA-List, a French public research institute, has been working for several years on real-time 3D localization applied to unloading and depalletizing. Its research confirms a simple fact: the more symmetrical or shiny an object is, the more difficult it is to localize, regardless of the technology used.

This challenge explains the rise of deep learning in this sector over the past few years. A well-trained deep learning model can generalize better than a fixed geometric rule when shapes, orientations, or shooting conditions vary significantly.

How does a 3D vision system work?

A 3D vision system involves five steps: image acquisition, point cloud reconstruction, object segmentation, pose calculation, and transmission of the coordinates to the robot. Each step depends on the previous one: an error earlier in the process affects the robotic arm.

• Acquisition. One or more 3D cameras capture the scene and generate a point cloud. On a bulk-unloading line, multiple viewing angles account for parts that overlap in the bin.
• Reconstruction. The software reconstructs the three-dimensional geometry of the observed scene. This step transforms a raw point cloud into a surface that can be used in subsequent steps.
• Segmentation. Objects of interest are isolated from the rest of the scene: bin, conveyor belt, or background. A deep learning model distinguishes an object from a shadow or a reflection.
• Pose calculation. An algorithm calculates the exact position and orientation of each object. This pose consists of six values: three for position and three for rotation.
• Transmission to the robot. The 3D coordinates are sent to the robot controller to guide the grasping motion. The robot adjusts its trajectory in a matter of milliseconds, without restarting the cycle.

Before these five steps are implemented in production, each camera must be calibrated: the exact position of the sensor relative to the scene and the robot is measured once and then recorded. An incorrect calibration will cause all subsequent coordinate data to be off, even if the rest of the production line is functioning perfectly.

Psycle’s Python SDK covers these five steps in a single framework. It is designed for teams that want to retain control over their deep learning models rather than rely on a proprietary black box.

What are the main 3D vision technologies?

There are four main categories of 3D vision technologies: stereoscopy, time-of-flight, structured light, and laser triangulation. Each involves a different trade-off between accuracy, speed, and working distance.

The choice almost always depends on the part being measured. A shiny metal part reflects structured light and skews the measurement; a time-of-flight sensor handles this type of surface better, at the cost of some accuracy, a trade-off frequently seen in certain lines of industrial sensors. Psycle systematically evaluates the customer’s actual sample before selecting a technology, rather than applying a standard solution.

This evaluation step often takes the form of an on-site test using the customer’s actual parts before committing to a specific technology. Psycle presented this approach at the SEPEM Industries Douai 2026 trade show, where several customer samples were tested live on different sensors.

3D vision and robot guidance: what are the industrial applications?

3D vision guides a robotic arm in three ways: it locates the part, calculates its grasping path, and then corrects its position in real time. These three functions are used in the food processing, automotive, and logistics industries.

• Part location: Locating a part whose position has not been calibrated in advance, such as a bottle that has slipped on a conveyor belt or a part that has been ejected from a mold without a fixed orientation.
• Trajectory calculation: determining the safest path to pick up the part without colliding with neighboring parts, a calculation that must remain compatible with the line’s production rate.
• Real-time correction: Adjust the robot’s motion if the part moves between detection and grasping, for example, due to conveyor vibration.

These three features eliminate the need for dedicated positioning tools for each product SKU, a direct benefit whenever a line processes multiple formats.

At Aretec, this approach made it possible to simplify the integration of AI-assisted machine vision into an existing production line without having to redesign the entire robotic cell. At Himber Technologies, the Psycle vision system was integrated into an existing workstation without causing any prolonged production downtime.

Unloading and bin picking: a major use case for industrial 3D vision

3D bulk sorting involves identifying and picking up loose parts from a bin without them being pre-arranged or oriented. This is the most common application of 3D vision in manufacturing.

The challenge does not come from the part itself, but from the disorder. The parts touch each other, overlap, and sometimes become deformed. A 2D vision system cannot distinguish between two stacked parts; a 3D system can, thanks to the pose calculation described above.

Certain materials further complicate the task: transparent materials allow the sensor’s light to pass through, flexible materials change shape between picks, and highly reflective materials create glare that resembles edges. Each case requires adjusting either the technology or the segmentation algorithm.

MCA Process has improved the de-bunking of unpackaged products with Psycle: parts without packaging, which are difficult to grasp without a clear visual reference point. In another customer case, Psycle reorganized a bin of nuggets to enable their automated retrieval. The same principle applies to mechanical parts, bottles, or flexible packaging.

3D vision vs. 2D vision: what are the differences?

2D vision measures a surface in terms of width and height. 3D vision adds depth. This third dimension changes everything as soon as a part is not lying flat in a known location.

2D vision remains useful for reading a barcode or inspecting a label on a flat surface: it is less expensive and processes images faster. It reaches its limits when it comes to locating a 3D object, measuring relief, or guiding a robot to a bulk item.

In practice, many production lines combine the two: a 2D camera for quality control at the end of the line, and a 3D camera upstream to guide the robot that feeds that same line. Each technology is used for the task where it is most effective, rather than forcing a single system to do everything.

Frequently Asked Questions about industrial 3D vision

What is industrial 3D vision?

Industrial 3D vision encompasses technologies that capture the depth of a scene, in addition to its surface, to enable an automated system to measure, control, or guide a robot with precision. It relies on specialized cameras and 3D reconstruction algorithms applied in production environments.

What is the difference between 2D and 3D vision?

2D vision analyzes a flat image in terms of width and height, which is sufficient for reading a barcode or inspecting a surface. 3D vision adds depth, which is essential for locating a part in three dimensions, measuring a surface profile, or guiding a robot through a bin of loose parts.

What 3D vision technologies are available?

The four main categories are stereoscopy, structured light, time-of-flight (ToF), and laser triangulation. Each offers a different balance between accuracy, speed, and working distance, and the choice depends on the intended industrial application.

How does 3D vision improve robot guidance?

It enables the robot to locate a part whose position is not known in advance, for example, a part loose in a bin, and then calculate a suitable grasping path in real time, without the need for dedicated positioning tools.