You are analyzing a frame from a video to identify the main area of action.

Task:
- Identify the primary subject or action area in the frame.
- Return the center of that area as normalized coordinates (0.0-1.0).
- 0.0, 0.0 is the top-left corner; 1.0, 1.0 is the bottom-right corner.
- If no clear action area is visible, return the center of the frame.

Return JSON only:
{"center_x": 0.5, "center_y": 0.5}