OUTCOMES: https://drive.google.com/drive/folders/1NxSlHTAYt0rh59rorZ9S8n7ySofK6pf2?usp=drive_link
Code Submission
Due to environment configuration issues on my new computer, I was unable to use Git for code submission. Instead, I have saved the complete implementation as a challenge.ipynb Jupyter notebook file for your review.
Code Explanation
PART 1 Solution Features:
Uses adaptive thresholding to handle variations in grassy backgrounds； Applies morphological operations to fill gaps in shapes； Filters noise based on contour area； Uses contour approximation to identify shape types； Calculates image moments to locate shape centers； Saves results to the specified directory； Optimization Approach & Key Methods
-
Color Space Conversion HSV Color Space: Switched from BGR to HSV for more accurate color detection； Precise Color Ranges: Defined specific HSV ranges for bright/fluorescent colors； Increased Saturation/Value Thresholds (150-255) to target vibrant shapes
-
Noise Reduction Strategies Morphological Operations: Applied closing and opening to clean up color masks； Area Filtering: Increased minimum area threshold to 2000 pixels to exclude grass patches； Circularity Check: Added circularity metric (>0.3) to filter out elongated grass shapes；
-
Shape Recognition Improvements Contour Approximation: Used polygon approximation for shape classification； Enhanced Circle Detection: Classified shapes with ≥8 sides as circles； Aspect Ratio Check: Differentiated squares from rectangles using bounding box ratios
-
User Experience Enhancements Auto-Close Display: Window automatically closes after 2 seconds； Clear Labeling: Combined color and shape names in annotations； Return Value: Added processed image return for potential further processing
-
Robustness Features Perimeter Validation: Added check to prevent division by zero； Error Handling: Included proper error messages for file loading issues； Directory Management: Automated output directory creation
Part 2
1.Solution Features (Optimized based on Problem 1): Modular Design: Frame processing logic is encapsulated in the independent process_frame() function； Stream Processing: Processes video frame by frame, simulating a real-time processing environment； Progress Feedback: Displays processing progress for monitoring long-running operations； Result Saving: Saves processed video to the specified directory； Real-time Preview: Optionally displays real-time processing results； Resource Management: Properly handles opening and releasing of video resources
- Optimizations and Improvements: Code Reuse: Problem 2 reuses the core processing logic from Problem 1； Performance Considerations: Added progress feedback in video processing to prevent UI freezing； User Experience: Added keyboard control (q key to quit) and real-time preview； Robustness: Added error checking and resource cleanup； These two solutions provide simple yet effective methods for shape detection tasks in both static images and videos, laying the foundation for more complex tasks to follow. The results will be saved to the specified "F:\airrobotics" directory.
Part 5 (ROS2 Integration) Status: I encountered memory allocation issues during ROS2 environment configuration that prevented implementation. After studying environment configuration methods and planning the implementation approach, I was unfortunately unable to resolve these hardware limitations in time for submission.
Part 6 (Enhanced Tracking) Concept: My approach for advanced object tracking involves implementing reinforcement learning through multiple path simulations. This method would better fit motion pattern recognition by training the system to predict object movement trajectories even during occlusion periods, significantly improving tracking consistency in challenging scenarios.