Extract Circle

## Extract Circle

First find points that meet the gradient response value, then use the RANSAC algorithm to fit a circle.

## Parameter Configuration Description

### Main Interface Parameters
![](/media/202601//1767687645.71788.png)

**[Gradient Threshold]** The grayscale gradient of the edge position along the scanning direction; points greater than this threshold are retained.
*   For clear images with obvious edges: Set a higher value (>40).
*   For blurry images or indistinct edges: Set a lower value (10-30).
![](/media/202601/tidu_1769762764.png)

**[Circle Scanning Direction]** Scan from inside out or from outside in, indicated by the arrow on the ROI.

**[Circle Polarity]**:
*   **Black to White:** Along the circular arrow, pointing from a low grayscale area to a high grayscale area.
*   **White to Black:** Along the circular arrow, pointing from a high grayscale area to a low grayscale area.
*   **All:** Can detect both black-to-white and white-to-black transitions.
![](/media/202602/2026-02-12_134405_0650450.041438465200031804.png)

**[Fitting Point Selection]**:
*   **All Points:** Use all detected edge points.
*   **Strongest Point:** Use the edge point with the maximum gradient value encountered along the scanning direction.
*   **First Row Points:** Use the first edge point encountered along the scanning direction.
*   **Last Row Points:** Use the last edge point encountered along the scanning direction.
![](/media/202602/2026-02-13_134437_9214050.23882749110783474.png)

**[Circle Filtering]**:
*   **Optimal Fit:** Comprehensively considers the best-fitted circle.
*   **Most Fitting Points:** The circle with the highest number of qualifying fitting points among all fitted circles.
*   **Strongest Fit:** The circle with the strongest gradient values for its fitting points among all fitted circles.
*   **First Circle:** The first circle fitted along the scanning direction.
*   **Last Circle:** The last circle fitted along the scanning direction.
*   **Specified ID Circle:** Filters out the circle with the specified ID along the scanning direction.

### Advanced Parameters
![](/media/202601//1767687645.9347115.png)

#### Filtering Parameters:

**[Gaussian Filter Threshold]** The σ in the Gaussian function, which can be understood as half the width of the transition region pixels. (A larger value reduces noise impact but blurs edges; a smaller value retains more detail but increases noise impact.)
![](/media/202602/滤波阈值_1770111631.png)

**[Sampling Points Count]** The number of uniform divisions of the ROI along the 360° direction.
![](/media/202602/2026-02-13_145748_1363740.1287834098883731.png)

**[Sampling Point Width]** Based on the center points divided evenly by the sampling points count, take the sampling point width to form small rectangular regions. Calculate and judge along the scanning direction within these small rectangular regions to check for qualifying edge points.
![](/media/202602/2026-02-13_150325_9327130.9314450900560899.png)

**[Minimum Fitting Points Count]** The minimum number of fitting points required for circle fitting. If the count is less than this value, the fitting is considered failed.
![](/media/202602/2026-02-13_151349_7629710.6790323035924312.png)

**[Distance Threshold]** Edge response points whose perpendicular distance to the fitted circle is less than this value are considered fitting points (also known as inliers). As shown in the figure below, edge points greater than the set distance threshold are displayed in red.
![](/media/202602/2026-02-13_151438_4592480.33066702526903646.png)

**[Maximum Iterations]** The maximum number of iterations required. If a circle meeting the conditions is fitted early, the fitting process can stop prematurely.

**[Iterations Without Improvement]** Stop iteration if the circle fitting effect does not continue to improve during the fitting process.

**[Score Threshold]** The final score used to judge the circle. If the score is less than this threshold, the fitting is considered failed. The range is 0-1. When all fitting points lie exactly on the fitted circle, the score is 1.

#### Judgment Parameters:

**[Size Conversion]** Two options: Scale and Calibration, used to calculate the circle's diameter.
*   If **Calibration** is selected, an additional port will be added to import the calibration file.
*   If **Scale** is selected, enter the scale value in the subsequent box.

**[Diameter Compensation]** Used to compensate for the diameter of the fitted circle.

**[Diameter Filtering]** When this judgment condition is enabled, if the diameter of the fitted circle is not within the specified range, the fitting is considered failed.

### ROI Settings
![](/media/202601//1767687646.1691122.png)

**[ROI Type]** The default ROI type for the detection area is an Annulus (Ring), but a Sector Ring can also be selected.

**[ROI Source]**
*   **Manual ROI:** Allows manual dragging of the ROI position and size in the window.
*   **Referenced ROI:** Allows connecting to an external ROI via a port.

**[Position Correction]** Selecting pose correction allows connection to a coordinate system, enabling the ROI area to move with the coordinate system.

**[Position Adjustment]** Adjust the position and size of the ROI.

## Output Description
*   **Run Result:** Outputs `True` if successful; outputs `False` if failed.
*   **Fitting Error:** The average error of the edge points used for circle fitting relative to the circle.
*   **Fitting Count:** The number of edge points used for circle fitting.
*   **Diameter Output:** The diameter of the detected circle.
*   **Center Output:** The center point of the detected circle.
*   **Circle Output:** The detected circle.
*   **Annular ROI:** The ROI used for detection.
*   **Fitting Points Array:** The set of edge points used for circle fitting.
![](/media/202602/2026-02-13_151740_1779330.27590874216226857.png)

---

Click here to jump to the corresponding Chinese manual page: [Extract Circle](/doc/86)