label_detection

A ROS package that detects flat objects in an image using affine invariant feature matching

Dependencies

affine_invariant_features

• https://github.com/yoshito-n-students/affine_invariant_features

object_detection_msgs

• https://github.com/yoshito-n-students/object_detection_msgs

Subscribed Topics

image_raw (sensor_msgs/Image)

Published Topics

labels_out (object_detection_msgs/Objects)

image_out (sensor_msgs/Image)

• original image on which labels are detected
• advertised and published when ~republish_image is true
• subtopics supported by image_transport are also published

Parameters

~reference_directory (string, default: "reference")

• path to directory which contains reference feature files (usually <label_name>.yml or <label_name>.yml.gz)

~parameter_file (string, default: "parameter.yml")

• path to file which specifies feature detection algorithm (usually <feature_name>.yml)
• must be the same file used in generating reference feature files

~match_stripes (double, default: -1.0)

• number of parallel threads matching features
• -1.0 means as many as possible

~match_ratio (double, default: 0.05)

• highly depends on type of feature detection algorithm
• compared to <number of matched features>/<number of all features in reference>
• label is detected if ~match_ratio is lower

~area_ratio (double, default: 0.1)

• compared to <area of label within image>/<total area of image>
• label is detected if ~area_ratio is lower

~desired_encoding (string, default: "bgr8")

• desired image encoding for internal processing
• the encoding of an incomming image will be converted to ~desired_encoding
• must be the same as the encoding of reference images

~republish_image (bool, default: false)

• republish an image when labels on it are detected
• useful for smaller ~queue_size of external time synchronizer for image and label messages

~image_transport (string, default: "raw")

• transport type of the subscribed image topic

Examples

label_detection_example

• https://github.com/yoshito-n-students/label_detection_example