Multi-Camera Alignment

## Multi-Camera Alignment

The positions of incoming material and bonding are both not fixed, typically using upper and lower cameras for dual-plane alignment and correction. Generally applied when the material is too large for the camera to capture the extraction target feature points, multiple cameras are selected for alignment.

## Usage Scenarios
Machine vision alignment refers to the process of using a machine vision system (camera, lens, light source, image processing software) to accurately identify and measure the position and angle (pose) of a target object (workpiece), and guide, adjust, or align it to a preset expected position and angle. Its core goal is to eliminate position and angle deviations to achieve high-precision positioning and assembly.
Simply put, the vision system tells the control system: The target is here now (Coordinates X, Y, Angle θ), and it differs from where it should be (Target Coordinates X0, Y0, Angle θ0) by (ΔX, ΔY, Δθ), please move this much to align.
Alignment is generally used to align objects on two planes, commonly used in film sticking and labeling. It involves situations where information captured by two cameras needs to be mutually converted.

![](/media/202601/2026-01-16_103750_4119460.3507559082787539.png)
#### Difference between Alignment and Correction
Compared to correction, alignment involves two planes and requires a mapping to convert the two planes into one plane.

Alignment:

![](/media/202603/2026-03-13_115204_0874040.13453579055629583.png)

Correction:

![](/media/202603/2026-03-13_115215_8866080.006531376542998624.png)

## Module Principle

![](/media/202603/2026-03-13_115204_0874040.13453579055629583.png)

![](/media/202603/2026-03-13_115347_3079510.48773395851870727.png)
The multi-camera alignment process consists of multiple pairs of cameras. The image data from one camera is transferred to the calibrated camera via mapping, and then a correction calculation is performed. That is, both are converted to mechanical points and then displacement compensation is calculated. Therefore, calibration and mapping need to be done in advance before alignment.
Multi-camera alignment only has the multi-point form. Each camera acquires one point, and each camera is calibrated with the mechanical equipment. Note that the order of input calibration and mapping needs to be the same as the order of start points. (The order of the calibration group/mapping group is the order of calibration reference; if calibration is less than the number of cameras, it will fail to run.)
The start point of multi-camera alignment is the camera doing the calibration, and the arrival point is the camera with only mapping.

Mechanical models are divided into XYR (with angle) and XY types (without angle). The calculation methods for the two types of alignment are different. Alignment with an angle requires the input calibration data to also be of the XYR type with an angle. Alignment without an angle has no requirement for calibration.
There are three alignment methods: point pair, point plus angle, and point plus line. Among them, the point pair uses the decomposition algorithm. Point plus angle and point plus line use the same algorithm; the line input will extract the line's angle and substitute it into the point plus angle algorithm to calculate the result.
### Compensation Amount
For the compensation amount in a mechanical model with angle R, the compensated R will affect the XY values. (That is, if only R is compensated, the XY position where the robot arm picks the workpiece will not change, but the actual mechanical XY coordinates will change slightly to compensate for the rotation.) The compensation amount will be added to the result after the last alignment ends. If the alignment has not ended and needs to continue, the result output will not add compensation. The compensation amount will not affect the judgment of the alignment completion accuracy of the correction.

![](/media/202603/2026-03-13_115409_8651340.6335112587194175.png)

Camera Start Shooting Pose: The fixed position when the machine picks up the workpiece (film in film sticking alignment) for shooting, generally the shooting position of the bottom camera. When the machine type has an angle R, the shooting pose has a significant impact on the result.
Alignment accuracy is the mechanical accuracy. When the alignment output is less than the alignment accuracy, the alignment is successful. Otherwise, if the number of alignments is less than the set number, error code 1 will be output, and alignment will continue.
The mechanical range is the movable range of the mechanism. When the output actual coordinate exceeds the range, error code 5 will be output.
## Usage Method

### Pre/Post-Sequence Modules

Camera image acquisition, feature matching, circle extraction are used to obtain points or angles for alignment. (Optimized feature matching is used to provide position correction for circle extraction.)
The start camera and arrival camera acquire the start image point and arrival image point respectively. Each camera acquires one point. When one camera acquires multiple points, **Calibration Reference** Mapping Reference needs to repeatedly reference the calibration.
Finally, all acquired data are aggregated to the multi-camera alignment, converted to mechanical points for alignment calculation.

## Parameter Configuration Description
【Camera Count】 The total number of cameras used for alignment.
#### Advanced Parameters
【Shooting Teach Pose】 The mechanical end coordinate when the camera can capture the material during the teaching process.
【Alignment Accuracy】 Mechanical coordinates, the accuracy range that alignment needs to reach.
【Correction Times】 The maximum number of times alignment is allowed. If the alignment result does not reach the alignment accuracy and is still within the mechanical travel range, alignment continues until the correction times.
【Mechanical Travel】 Allowed travel range of the robot arm.
【Compensation Amount】 Mechanical coordinates, compensated in the mechanical coordinate system. When clicking subscribe, the compensation amount can be introduced via the port.
【Reference Image】 The image used as a reference during processing.

## Output Description

![](/media/202603/2026-03-13_115517_3566590.263618126443544.png)
Accumulated Offset is the result of accumulating multiple alignments when the maximum number of alignments is not 1. It is the sum of the mechanical alignment displacements (compensation will be added).
The output has two sets: Pick Pose and Offset Pose. Pick Pose is the actual picking pose of the robot arm, which is also the judgment range for [Mechanical Travel]. Offset Pose is the offset of the robot arm relative to the teaching position and does not judge mechanical travel.
| Feature | Pick Pose | Offset Pose |
|---|---|---|
| Coordinate Type | Absolute Coordinate (GPS style) | Relative Coordinate (Reference object style) |
| Safety Check | Full check of mechanical travel | No range check |
| Main Function | Final execution position | Position fine-tuning |
| Analogy | Detailed delivery address | "Front left of you" |
| Dependency | Absolute Coordinate (GPS style) | Contains teach coordinates + offset based only on teach coordinates |
## Precautions
1. The forward operation when referencing mapping is from the start camera point to the arrival camera point. However, in the integrated Dual Camera Alignment and Multi-Camera Alignment, the mapping uses the inverse operation. That is, it converts the point on the arrival camera to the point on the start camera through mapping, and then uses the calibration of the start camera for conversion. If the actual model requires the forward operation of mapping, the user needs to build the flow themselves and use mechanical point alignment for calculation.

![](/media/202603/2026-03-13_115530_4129330.9279434095048543.png)

## Problem Finding
1. When alignment is inaccurate, pay attention to whether the feature matching module acquires an accurate image.
2. If still unable to stabilize correction, test the accuracy of calibration and mapping.
(Calibration Mapping Problem Test Method:
Change the picking shooting position by 2mm, observe if the result output is offset by 2mm. If the result is incorrect in this case, it indicates a calibration error.
Change the bonding shooting position by 2mm, observe if the result output is offset by 2mm. If the result is incorrect in this case, it indicates a mapping error.)
3. If there is a large offset angle leading to inaccurate XY, it may be caused by an error in the calculation of the calibration rotation center. Try to recalibrate, modify the calibration method, and improve the calculation of the rotation center (add calibration verification or use the long radius calibration module).
4. If the correction is accurate in the center of the field of view but there is XY deviation around the periphery, consider image distortion (use the Image Distortion Correction module), or calibration scale calculation error (use the Calibration Correction module).

---

Click here to jump to the corresponding Chinese manual page: [Multi-Camera Alignment](/doc/149)