[Documentation] [TitleIndex] [WordIndex

Package Summary

The anj_featurenav package provides a learning jockey and a navigating jockey for the Large Maps framework (LaMa). It learns a path by saving image features and is able to follow the same path. It is based on algorithms provided by OpenCV (free ones).

Overview

anj_featurenav implements a feature-based learning and navigating jockey based on free OpenCV feature descriptor and matcher. The package defines the feature extractor and the descriptor matcher functions required by the featurenav_base package to obtain a working couple learning/navigating jockeys.

Usage

ROS API

The words feature and descriptor are used as synonymous.

Parameters

General
~<name>/feature_detector/type (String, default: "FAST") ~<name>/descriptor_extractor/type (String, default: "BRIEF") ~<name>/descriptor_matcher/type (String, default: "FlannBased")
For the FAST algorithm
~<name>/feature_detector/threshold (Int, default: 1) ~<name>/feature_detector/nonmax_suppression (Bool, default: true)
For the STAR algorithm
~<name>/feature_detector/max_size (Int, default: 16) ~<name>/feature_detector/response_threshold (Int, default: 30) ~<name>/feature_detector/line_threshold_projected (Int, default: 10) ~<name>/feature_detector/line_threshold_binarized (Int, default: 8) ~<name>/feature_detector/suppress_nonmax_size (Int, default: 5)
For the ORB algorithm, as feature detector
~<name>/feature_detector/scale_factor (Float, default: 1.2) ~<name>/feature_detector/n_features (Int, default: 500) ~<name>/feature_detector/n_levels (Int, default: 8) ~<name>/feature_detector/edge_threshold (Int, default: 31) ~<name>/feature_detector/first_level (Int, default: 0) ~<name>/feature_detector/wta_k (Int, default: 2) ~<name>/feature_detector/score_type (Int, default: 0 (i.e. Harris score)) ~<name>/feature_detector/patch_size (Int, default: 31)
For the MSER algorithm
~<name>/feature_detector/delta (Int, default: 5) ~<name>/feature_detector/min_area (Int, default: 60) ~<name>/feature_detector/max_area (Int, default: 14400) ~<name>/feature_detector/max_variation (Float, default: 0.25) ~<name>/feature_detector/min_diversity (Float, default: 0.2) ~<name>/feature_detector/max_evolution (Int, default: 200) ~<name>/feature_detector/area_threshold (Float, default: 1.01) ~<name>/feature_detector/min_margin (Float, default: 0.003) ~<name>/feature_detector/edge_blur_size (Int, default: 5)
For the GFTT algorithm
~<name>/feature_detector/max_corners (Int, default: 1000) ~<name>/feature_detector/block_size (Int, default: 3) ~<name>/feature_detector/quality_level (Float, default: 0.01) ~<name>/feature_detector/min_distance (Float, default: 1) ~<name>/feature_detector/k (Float, default: 0.04) ~<name>/feature_detector/use_harris_detector (Bool, default: false)
For the Dense algorithm
~<name>/feature_detector/feature_scale_levels (Int, default: 1) ~<name>/feature_detector/init_xy_step (Int, default: 6) ~<name>/feature_detector/init_img_bound (Int, default: 0) ~<name>/feature_detector/init_feature_scale (Float, default: 1) ~<name>/feature_detector/feature_scale_mul (Float, default: 0.1) ~<name>/feature_detector/vary_xy_step_with_scale (Bool, default: true) ~<name>/feature_detector/vary_img_bound_with_scale (Bool, default: false)
For the ORB algorithm, as descriptor extractor
~<name>/descriptor_extractor/scale_factor (Float, default: 1.2) ~<name>/descriptor_extractor/n_features (Int, default: 500) ~<name>/descriptor_extractor/n_levels (Int, default: 8) ~<name>/descriptor_extractor/edge_threshold (Int, default: 31) ~<name>/descriptor_extractor/first_level (Int, default: 0) ~<name>/descriptor_extractor/wta_k (Int, default: 2) ~<name>/descriptor_extractor/score_type (Int, default: 0 (i.e. Harris score)) ~<name>/descriptor_extractor/patch_size (Int, default: 31)
For the BRIEF algorithm
~<name>/descriptor_extractor/bytes (Int, default: 32)
For the BruteForce algorithm
~<name>/descriptor_matcher/norm (String, default: "L2") ~<name>/descriptor_matcher/cross_check (Bool, default: false)

The SimpleBlog algorithm takes no parameter, cf. http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=simpleblob#simpleblobdetector.

The FlannBased algorithm takes no parameter, cf. http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html?highlight=flannbasedmatcher#bfmatcher.


2019-10-19 12:30