Color Recognition

1. Experiment Objective

Use OpenCV to detect specified color regions in the HSV color space, with support for ROI mouse color sampling, dynamic HSV tuning, and morphological filtering.

2. Experiment Principle

1. Key Terms

2. Technical Architecture

Terminal 1: start agent + camera + joint models
Terminal 2: detect_color_node (HSV threshold segmentation → morphological filtering → contour detection → OpenCV visualization)
        supports ROS2 dynamic parameter tuning + ROI mouse color sampling + HSV threshold file persistence

Data Flow: Image subscription → image conversion → BGR→HSV → cv2.inRange threshold segmentation → morphological filtering → cv2.findContours → area/radius filtering → draw enclosing circles + labels → real-time display.

3. Experiment Steps

1. Hardware Preparation

2. Start Agent + Camera + Joint Model (Terminal 1)

xxxxxxxxxx
ros2 launch microros_v2_bringup car_base.launch.py

3. Start Color Recognition (Terminal 2)

xxxxxxxxxx
ros2 launch microros_v2_opencv_visual_pkg detect_color.launch.py

4. Operation Instructions

Two operation modes:

ROI color sampling flow: Drag a mouse box to select the region → compute the min/max of each channel inside the ROI → subtract the expansion amount for the lower bound and add it for the upper bound → S/V upper bounds are kept no lower than 253 → automatically saved to color_hsv.txt → switch back to detect mode.

Dynamic parameter tuning:

xxxxxxxxxx
ros2 param list /detect_color_node # list all parameters currently "declared" by the /detect_color_node node
ros2 param set /detect_color_node min_target_area 100.0 # at runtime, change min_target_area to 100.0 (unit: pixel^2, i.e. the contour area)

Adjustable parameters: HSV upper/lower bounds, morphological kernel size, minimum area/radius, window flip/rotation, etc.

5. How to Stop

Press Ctrl+C to stop the terminals one by one.

4. Experiment Observations

• Default detection (red): Red objects are marked with a green circle and a red center dot, labeled Color 1
• Multiple targets: Multiple objects are marked simultaneously; the one with the largest area is the purple primary target, the rest are green
• ROI sampling switch: After box-selecting a new color, detection switches from red to the newly selected color, and the HSV thresholds update and save automatically
• Binary mask side by side (show_binary_window:=true): the binary mask is shown side by side on the right

5. Program Analysis

Source code paths:

• Launch: ~/microros_ws/src/microros_v2_opencv_visual_pkg/launch/detect_color.launch.py
• Node: ~/microros_ws/src/microros_v2_opencv_visual_pkg/microros_v2_opencv_visual_pkg/detect_color.py

1. Core Node Analysis

DetectColorNode uses HSV threshold segmentation + morphological filtering + contour detection, with support for ROI mouse color sampling and ROS2 dynamic parameter tuning.

​x
# Source file: ~/microros_ws/src/microros_v2_opencv_visual_pkg/microros_v2_opencv_visual_pkg/detect_color.py
class DetectColorNode(Node):
    def __init__(self):
        super().__init__('detect_color_node')
        # HSV threshold parameters (adjustable at runtime via ros2 param set)
        self.declare_parameter('Hmin', 0)
        self.declare_parameter('Hmax', 9)
        self.declare_parameter('Smin', 85)
        self.declare_parameter('Smax', 253)
        self.declare_parameter('Vmin', 126)
        self.declare_parameter('Vmax', 253)
        self.declare_parameter('min_target_area', 60.0)
        self.declare_parameter('min_target_radius', 5.0)
        self.declare_parameter('morph_kernel_size', 5)
        self.declare_parameter('show_binary_window', False)
​
    def detect_color_targets(self, hsv):
        lower = np.array([self.Hmin, self.Smin, self.Vmin])
        upper = np.array([self.Hmax, self.Smax, self.Vmax])
        mask = cv.inRange(hsv, lower, upper)
​
        # Morphological filtering: opening removes noise → closing fills holes
        kernel = cv.getStructuringElement(cv.MORPH_RECT, (ksize, ksize))
        mask = cv.morphologyEx(mask, cv.MORPH_OPEN, kernel)
        mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, kernel)
​
        # Contour detection + area/radius filtering
        contours, _ = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
        targets = []
        for cnt in contours:
            area = cv.contourArea(cnt)
            if area < self.min_target_area:
                continue
            (x, y), radius = cv.minEnclosingCircle(cnt)
            if radius < self.min_target_radius:
                continue
            targets.append((cnt, area, (int(x), int(y)), int(radius)))
        return targets

Key logic notes:

• Red is detected by default: Hmin=0, Hmax=9 covers red at both ends of the HSV hue wheel (0°~9°)
• Morphological filtering order: opening first (remove isolated noise), then closing (fill holes inside the target); the two operations complement each other
• cv2.RETR_EXTERNAL: extracts only the outermost contours and ignores nested contours (reduces redundant detections)
• Dual area + radius filtering: min_target_area=60 filters out pixel-level noise, min_target_radius=5 filters out line-like artifacts
• S/V upper bounds are kept no lower than 253: after ROI sampling, the saturation/value ranges are stretched automatically for tolerance to lighting changes

2. Key Parameters

Parameter definition files:

• Launch: ~/microros_ws/src/microros_v2_opencv_visual_pkg/launch/detect_color.launch.py
• Node: ~/microros_ws/src/microros_v2_opencv_visual_pkg/microros_v2_opencv_visual_pkg/detect_color.py

3. Node Description

6. Frequently Asked Questions