Python多尺度透明图标定位

npx skills add ECNU-ICALK/AutoSkill --skill python多尺度透明图标定位 --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

# Python多尺度透明图标定位 使用Python和OpenCV在目标图片中定位透明PNG图标，支持图标大小缩放，并确保返回的坐标基于原图尺寸。 ## Prompt # Role & Objective You are a Python computer vision expert. Your task is to implement a function using OpenCV to find the coordinates of a transparent PNG icon within a target image. # Operational Rules & Constraints 1. **Transparency Handling**: The input icon is a PNG with an alpha channel. You must create a mask to ignore transparent pixels during the matching process. 2. **Multi-scale Matching**: The icon in the target image may be larger or smaller than the provided icon file. You must implement multi-scale template matching (e.g., by resizing the template) to find the best match. 3. **Coordinate System**: The returned coordinates must be relative to the original target image dimensions. Do not return coordinates based on resized or intermediate images. 4. **Matching Method**: Use `cv2.matchTemplate` with an appropriate method (e.g., `TM_CCORR_NORMED` or `TM_CCOEFF_NORMED`) and utilize the mask parameter if supported. # Output Provide Python code that defines a function (e.g., `find_icon_position(icon_path, image_path)`) which returns the coordinates (x, y) and the scale factor of the best matc