Feature Matching

>i **Template Description**
>
> 1. [xxx] Needs to be replaced
> 2. After document editing is completed, please delete the template description

## [Feature Matching]

Find similar graphic regions based on the contour feature information of the template image.

## Module Principle
Find similar graphic regions based on the contour feature information of the template image.

## Usage Scenarios
Find objects with certain contour features

## Usage Method

### Predecessor/Successor Modules
Generally used in conjunction with the Camera Image Acquisition module

### Usage Example
[Record operation video]

## Feature Template
[Some matching modules need to save feature templates in advance, which need to be introduced in detail in this section]

## Parameter Configuration Description

### Basic Parameters
【Matching Score】The similarity between the template and the target in the search image. The detected target will only be found when the similarity is greater than the set value. The default is 0.7.
【Search Angle】Input the minimum and maximum search angles. Only targets within this angle range can be matched. The range is from -180 to 180.
【Display Settings】Display settings such as ROI region, graphic results, mask display, etc., including whether to display, color selection, font size, etc.
【Search Count】The maximum number of targets to be searched.
【Image Filtering】Image filtering is mainly used for noise suppression and edge enhancement in the preprocessing stage, with the aim of improving detection stability and matching accuracy.
【Positioning Mode】Precise Positioning: More accurate; Fast Positioning: Faster speed, not as accurate as precise positioning.
【Consider Polarity】Parameter used to control whether pixel intensity sign information (i.e., light-dark relationship) is retained in the feature descriptor during the matching process. The core meaning is: whether to allow the matching algorithm to distinguish between "light to dark" and "dark to light" edge change directions.
【ROI Type】The default ROI type for the detection area is Rotated Rectangle.
【ROI Source】Manual ROI: Can manually drag the position and size of ROI in the window; Reference ROI: Can connect external ROI at the port.
【Position Correction】After selecting pose correction, you can connect the coordinate system to make the ROI region move with the coordinate system.
【Position Adjustment】Adjust the position and size of ROI.
### Advanced Parameters
【Scaling Ratio】The scaling ratio defaults to 1.00~1.00 (i.e., matching targets without scaling). When matching scaled targets, this parameter needs to be adjusted according to the target scaling ratio. (This parameter is currently invalid)
【Min Target Center Distance】The minimum pixel distance that must be met between two adjacent feature points. If the distance between two feature points is less than this threshold, only the point with the higher response value is retained to reduce the number of feature points and reduce the complexity of subsequent matching calculations.
【Target Region Avoidance (Min Distance)】Configuration option to shield or exclude specific areas in the image. Its core purpose is to prevent feature points from being extracted or matched in irrelevant or interference areas (such as background, occlusions, motion blur areas), thereby improving matching accuracy and robustness.
【Recheck After Rotation (Angle)】
【Sorting Standard】Sort the matched results according to the standard
【Sorting Interval】
### Template Edit
【Auto Edge Gradient】After enabling "Auto Edge Gradient", the template will only retain clear, stable, and geometrically significant edge structures (such as object contours, corners, edges). Please prioritize using "Auto Edge Gradient". If fine-tuning is needed, then perform manual edge gradient setting.
【Auto Positioning Scale】Auto Positioning Scale: Algorithm intelligently selects, suitable for scenes with unknown scale changes, strong robustness. Manual Positioning Scale: User explicitly sets scale range or fixed value, but needs to predict target size.
【Detection Scale】Detection scale should be smaller than positioning scale [1,16]. Through actual testing of matching effects under different scale ranges, choose the parameter setting that achieves the best balance between accuracy and efficiency.
【Edge Color】Edge color of the model
【Update Image】After clicking, the image input port of this module will be read to the interface. After clicking Generate Model, the template image will be switched to the current image. This button needs to be clicked when switching model images.
【Position Adjustment】Specify adjusting model roi position, used for finer roi adjustment
【Generate Model】After clicking, generate a new model based on the current image and parameters

【Enable Mask】After checking Enable Mask, you can edit mask parameters. The core function of mask parameters is to precisely control the matching area, used to limit the effective area of feature detection and matching. It is a spatial constraint tool that allows users to specify which areas in the image participate in feature extraction and matching, and which areas are ignored.

【Manual】Manually draw and erase required masks through brushes, erasers, etc.

【Mask Brush】Brush: Add mask according to drawing;
【Brush Size】Select the diameter size of the brush and eraser
【Mask Eraser】Similar to eraser, erase mask according to drawing

【Auto】Automatically add masks according to conditions
【Auto Threshold】Quickly add masks according to color conformity
【Gray Threshold】Used in conjunction with auto threshold parameter, quickly add masks when conforming to gray range and auto threshold. For example, when auto mask selects threshold middle, those within the gray threshold range will be added with masks.

【Mask Shape】Quickly add mask according to the selected roi shape,
【Mask Shape】Different roi shapes such as rectangle, circle, etc.
【Add】After clicking, generate mask based on roi
【Position Adjustment】Adjust roi position

【Cover All】Add mask to the entire image
【Erase All】Erase all masks

【Enable Matching Verification】Template mask
【Model Generation】Click this button to regenerate verification model
【Edge Color】Verification model edge color
【Position Adjustment】Adjust verification roi position
【Auto Edge Gradient】
【Auto Positioning Scale】
【Detection Scale】
【Matching Score】
【Search Angle Range】
【Matching Result Angle】
【Search Region】

## Output Description

【Running Result】Output True if target is matched, otherwise False
【Target Count】Number of matched targets
【Matching Score】Matching score value corresponding to matching result
【Angle Offset】Angle of matching result
【Position Offset】Position offset of matching result relative to template
【Detection Angle】Angle of matching result
【Detection Position】Detection position corresponding to matching result
【Output Coordinate System】Coordinate system of matching result, constructed by detection position and detection angle
【Matching Score Array】Array of matching score values corresponding to matching results, array size is the same as matched target count
【Detection Point Array】Array of detection positions corresponding to matching results, array size is the same as matched target count
【Detection Angle Array】Array of detection angles corresponding to matching results, array size is the same as matched target count
【Reference Point Array】Array of offset reference points generated based on detection position, array size relates to the number of reference points added in model editing. If no reference points are added, this option is invalid
【Regular Rectangle ROI】Rectangular border drawn by the roi of this module
【Mask Region】Mask region edited by this module

## Other Supplements

## Precautions

---

Click here to jump to the corresponding Chinese manual page: [Feature Matching](/doc/76)