robokudo.annotators.cluster_color

Robokudo Color Analysis Module

This module provides functionality for semantic color analysis of object hypotheses. It analyzes RGB regions of interest (ROIs) and masks to determine dominant colors using HSV color space segmentation.

Warning

The non-numba version of color counting may contain bugs in color counting.

Classes

Color

Enumeration of semantic colors used for classification.

ClusterColorAnnotator

Calculate the semantic color for every Object Hypothesis (cluster) that has a RGB ROI and a Mask.

Module Contents

class robokudo.annotators.cluster_color.Color

Bases: enum.Enum

Enumeration of semantic colors used for classification.

  • Primary colors: RED, GREEN, BLUE

  • Secondary colors: YELLOW, CYAN, MAGENTA

  • Grayscale: WHITE, BLACK, GREY

RED = (0,)
YELLOW = (1,)
GREEN = (2,)
CYAN = (3,)
BLUE = (4,)
MAGENTA = (5,)
WHITE = (6,)
BLACK = (7,)
GREY = (8,)
class robokudo.annotators.cluster_color.ClusterColorAnnotator(name='ClusterColorAnnotator', descriptor=Descriptor())

Bases: robokudo.annotators.core.BaseAnnotator

Calculate the semantic color for every Object Hypothesis (cluster) that has a RGB ROI and a Mask.

This annotator analyzes object hypotheses by:

  • Converting RGB image regions to HSV color space

  • Using HSV thresholds to classify pixels into semantic color categories

  • Counting pixel distributions for each color

  • Annotating objects with their dominant colors

The color classification uses configurable thresholds for:

  • Hue ranges for primary/secondary colors

  • Saturation threshold for chromatic colors

  • Value thresholds for black/white/grey

class Descriptor

Bases: robokudo.annotators.core.BaseAnnotator.Descriptor

Configuration descriptor for the ClusterColorAnnotator.

Defines parameters that control the color analysis behavior:

  • Number of color divisions in hue space

  • Thresholds for color classification

  • Minimum ratio for color annotation

class Parameters

Parameters for color analysis configuration.

num_of_colors = 6
color_range = 42.666666666666664
jit_compile_on_init = False
min_value_color = 60
min_saturation_color = 60
max_value_black = 60
min_value_white = 120
ratio_annotation_threshold = 0.2
analysis_scope
parameters
color_name_to_bgr_values
color_hue_positions
cluster_color_info = []
cluster_rois = []
count_colors_numba(hsv_image: numpy.typing.NDArray, mask: numpy.typing.NDArray)

Wrapper function for the numba version of the count colors method.

The wrapper handles the pythonic/object style data and prepares everything for the requirements of the numba method.

Parameters:

  • hsv_image (numpy.ndarray) – Image in HSV color space

  • mask (numpy.ndarray) – Binary mask indicating pixels to analyze

Returns:

A tuple with the number of analyzed pixels (given by the mask) as well as a dict with the counted colors. The key is a Color enum from this module.

Return type:

tuple(int, dict)

static count_colors_numba_impl(hsv_image: numpy.typing.NDArray, mask: numpy.typing.NDArray, total_amount_of_colors: int, min_saturation_color: int, min_value_color: int, max_value_black: int, min_value_white: int, number_of_colors: int, color_range: float)

Iterate over the masked pixels on a given hsv image to count how many pixels can be assigned to the defined color ranges. Requires the numba library and passes the required parameters to avoid class access and object-mode from numba.

Parameters:

  • hsv_image (numpy.ndarray) – Image in HSV color space

  • mask (numpy.ndarray) – Binary mask indicating pixels to analyze

  • total_amount_of_colors (int) – Total number of color categories

  • min_saturation_color (int) – Minimum saturation for chromatic colors

  • min_value_color (int) – Minimum value for chromatic colors

  • max_value_black (int) – Maximum value for black classification

  • min_value_white (int) – Minimum value for white classification

  • number_of_colors (int) – Number of hue divisions

  • color_range (float) – Size of each hue division

Returns:

A tuple with the number of analyzed pixels (given by the mask) as well as a numpy array with the counted colors. The id refers the indices of the Color enum from this module.

Return type:

tuple(int, numpy.ndarray)

count_colors(hsv_image: numpy.typing.NDArray, mask: numpy.typing.NDArray)

Non-numba version of the count_colors_numba_impl method. Use only if you can’t use numba.

Warning

Seems to contain a bug in color counting.

Parameters:

  • hsv_image (numpy.ndarray) – Image in HSV color space

  • mask (numpy.ndarray) – Binary mask indicating pixels to analyze

Returns:

A tuple with the number of analyzed pixels and a dict with color counts

Return type:

tuple(int, dict)

update()

Process current scene to analyze colors of object hypotheses.

Steps:

  • Gets current point cloud and color image

  • Creates color annotations for each object hypothesis

  • Updates visualization with color information

  • Prepares 3D visualization geometries

Returns:

SUCCESS after processing is complete

Return type:

py_trees.Status

create_color_annotations(color)

Create color annotations for all object hypotheses.

For each object hypothesis with a valid ROI mask:

  • Extract color image region and mask

  • Convert to HSV color space

  • Count pixel colors using numba-optimized implementation

  • Sort colors by frequency

  • Add color annotations for colors above ratio threshold

  • Store color information for visualization

Parameters:

color (numpy.ndarray) – Input color image

draw_visualization(visualization_img)

Draw visualization of detected colors on the image.

For each ROI:

  • Draw bounding box in dominant color

  • Add text label with ROI index and dominant color name

  • Create color histogram showing distribution of detected colors

Parameters:

visualization_img (numpy.ndarray) – Image to draw visualization on

Returns:

Image with visualization overlays

Return type:

numpy.ndarray