如何改进图像处理校准？添加HSV/阈值滑动条实现无人机形状检测

Got it, let's build a robust shape detection system with trackbar-based calibration tailored for your drone footage needs. Your existing code struggles with non-ideal lighting, so adding real-time HSV and threshold controls will let you fine-tune parameters on the fly. Here's a complete, optimized implementation:

import argparse
import cv2
import time
import datetime

# Initialize global variables for trackbar values
h_low, h_high = 0, 179
s_low, s_high = 0, 255
v_low, v_high = 0, 255
threshold_val = 127

# Empty callback function for trackbars (required by OpenCV)
def nothing(x):
    pass

def detect_shape(contour):
    # Approximate contour to simplify polygon edges
    peri = cv2.arcLength(contour, True)
    approx = cv2.approxPolyDP(contour, 0.04 * peri, True)
    
    # Classify shape based on vertex count
    if len(approx) == 3:
        return "Triangle"
    elif len(approx) == 4:
        # Distinguish square from rectangle using aspect ratio
        x, y, w, h = cv2.boundingRect(approx)
        aspect_ratio = float(w) / h
        if 0.95 <= aspect_ratio <= 1.05:
            return "Square"
        else:
            return "Rectangle"
    elif len(approx) > 4:
        return "Circle"
    else:
        return "Custom Shape"  # Extend this for your specific target

def main():
    # Parse command line arguments for video input
    ap = argparse.ArgumentParser()
    ap.add_argument("-v", "--video", help="path to drone video file")
    args = vars(ap.parse_args())

    # Initialize video capture: use webcam if no video path provided
    if args.get("video", None) is None:
        cap = cv2.VideoCapture(0)
        time.sleep(2.0)  # Warm up camera
    else:
        cap = cv2.VideoCapture(args["video"])

    # Create windows for trackbars and output visualization
    cv2.namedWindow("Trackbars", cv2.WINDOW_NORMAL)
    cv2.namedWindow("Original", cv2.WINDOW_NORMAL)
    cv2.namedWindow("Mask", cv2.WINDOW_NORMAL)

    # Create trackbars for HSV range and threshold calibration
    cv2.createTrackbar("H Low", "Trackbars", h_low, 179, nothing)
    cv2.createTrackbar("H High", "Trackbars", h_high, 179, nothing)
    cv2.createTrackbar("S Low", "Trackbars", s_low, 255, nothing)
    cv2.createTrackbar("S High", "Trackbars", s_high, 255, nothing)
    cv2.createTrackbar("V Low", "Trackbars", v_low, 255, nothing)
    cv2.createTrackbar("V High", "Trackbars", v_high, 255, nothing)
    cv2.createTrackbar("Threshold", "Trackbars", threshold_val, 255, nothing)

    while True:
        ret, frame = cap.read()
        if not ret:
            break  # Exit loop if video ends

        # Resize frame for better visibility (adjust based on drone resolution)
        frame = cv2.resize(frame, (640, 480))
        # Reduce noise with Gaussian blur (critical for aerial footage)
        blurred = cv2.GaussianBlur(frame, (5, 5), 0)
        # Convert to HSV color space (more lighting-resistant than RGB)
        hsv = cv2.cvtColor(blurred, cv2.COLOR_BGR2HSV)

        # Get current trackbar values
        h_low = cv2.getTrackbarPos("H Low", "Trackbars")
        h_high = cv2.getTrackbarPos("H High", "Trackbars")
        s_low = cv2.getTrackbarPos("S Low", "Trackbars")
        s_high = cv2.getTrackbarPos("S High", "Trackbars")
        v_low = cv2.getTrackbarPos("V Low", "Trackbars")
        v_high = cv2.getTrackbarPos("V High", "Trackbars")
        threshold_val = cv2.getTrackbarPos("Threshold", "Trackbars")

        # Create HSV mask and apply binary threshold
        lower_hsv = (h_low, s_low, v_low)
        upper_hsv = (h_high, s_high, v_high)
        mask = cv2.inRange(hsv, lower_hsv, upper_hsv)
        _, thresh = cv2.threshold(mask, threshold_val, 255, cv2.THRESH_BINARY)

        # Find contours in the processed mask
        contours, _ = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        # Process each detected contour
        for c in contours:
            # Filter out tiny contours (adjust based on 100ft target size)
            area = cv2.contourArea(c)
            if area < 500:  # Lower this if your target is smaller at 100ft
                continue

            # Identify the shape
            shape = detect_shape(c)

            # Draw contour and label on original frame
            M = cv2.moments(c)
            if M["m00"] != 0:
                cX = int(M["m10"] / M["m00"])
                cY = int(M["m01"] / M["m00"])
                cv2.drawContours(frame, [c], -1, (0, 255, 0), 2)
                cv2.putText(frame, shape, (cX - 50, cY), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

                # Capture timestamped image when shape is detected
                timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
                capture_path = f"shape_capture_{timestamp}.jpg"
                cv2.imwrite(capture_path, frame)
                print(f"Captured target: {capture_path}")

        # Display output windows
        cv2.imshow("Original", frame)
        cv2.imshow("Mask", thresh)

        # Exit on 'q' key press
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break

    # Cleanup resources
    cap.release()
    cv2.destroyAllWindows()

if __name__ == "__main__":
    main()

Key Features Breakdown #

• Real-Time Trackbar Calibration: The Trackbars window lets you tweak HSV ranges and threshold values live, so you can adapt to cloudy, sunny, or shaded conditions instantly.
• Drone-Optimized Preprocessing: Gaussian blur cuts down on aerial noise, and HSV color space handles lighting shifts better than standard RGB.
• Adaptive Shape Detection: The detect_shape function uses contour approximation to identify common shapes—extend it by adding vertex checks for your custom target.
• Automated Image Capture: Timestamped filenames ensure you never overwrite captures, and the system saves frames only when valid shapes are detected.
• Adjustable Contour Filtering: The area threshold (500) is set for 100ft targets—lower it if your shape is smaller, or raise it to filter out noise.

How to Use #

1. Run the Code: Use python shape_detector.py for webcam testing, or python shape_detector.py --video path/to/drone_footage.mp4 to analyze recorded video.
2. Calibrate: Adjust trackbars until the Mask window shows only your target shape (no background clutter).
3. Capture: The system automatically saves frames when shapes are detected—check your working directory for the files.
4. Customize: Modify the detect_shape function to add support for your specific custom shape, or tweak the contour area threshold for your target size.

Troubleshooting Tips #

• Small Targets: If your shape is hard to detect at 100ft, reduce the area < 500 threshold or increase the video resolution.
• Flickering Mask: Widen the HSV ranges slightly or add a morphological operation like cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, (3,3)) to clean up noise.
• Jittery Contours: Add a frame buffer to average contour positions over time, or use cv2.boundingRect to track the target more smoothly.

内容的提问来源于stack exchange，提问作者Andromedan