wass_match configuration

To adapt the feature matching procedure to the characteristics of the processed stereo data, the wass_match executable can be configured with a dedicated file whose name and location is defined in <WASS_ROOT>/WASSjs/settings.json (See the pipeline configuration page for more info).

By convention, the wass_match configuration file is named matcher_config.txt and is placed in the "configuration folder" containing the calibration data of the currently processed stereo dataset. To generate a new configuration file with the default values, run the wass_match executable with the option --genconfig:

cd <WASS_ROOT>/dist/bin
./wass_match --genconfig

wass_match  v. 1.0_heads/master-0-g5a7e63d
---------------------------------------
 Darwin-15.5.0 - AppleClang

(wass_match) info: Generating matcher_config.txt ...
(wass_match) info: Done!

The generated file should be similar to the following:

# OpenSURF Hessian threshold
# 
#FEATURE_HESSIAN_THRESHOLD=0.0001

# OpenSURF init samples
# 
#FEATURE_INIT_SAMPLES=1

# Minimum distance allowed between two features (in px)
# 
#FEATURE_MIN_DISTANCE=10

# OpenSURF number of layers
# 
#FEATURE_N_LAYERS=4

# OpenSURF number of octaves
# 
#FEATURE_N_OCTAVES=4

# Matcher payoff lambda
# 
#MATCHER_LAMBDA=1e-05

# Max matches epipolar distance
# 
#MATCHER_MAX_EPI_DISTANCE=0.5

# Matcher maximum number of rounds to perform
# 
#MATCHER_MAX_ROUNDS=20

# Matcher minimum required group size
# 
#MATCHER_MIN_GROUP_SIZE=5

# Matcher population threshold
# 
#MATCHER_POPULATION_THRESHOLD=0.7

# Maxmum number of image features to extract
# 
#NUM_FEATURES_PER_IMAGE=2000

Each configuration option is composed by a key=value pair that can be freely edited by the user. Lines beginning with a # are considered as comments and hence are discarded by wass_match (ie. the default value is assumed).

Here is a brief explanation of each configuration option:

key	Value
FEATURE_HESSIAN_THRESHOLD	Hessian threshold used by OpenSURF feature extractor
FEATURE_INIT_SAMPLES	"init samples" parameter used by OpenSURF feature extractor
FEATURE_MIN_DISTANCE	Minimum required distance (in pixel) between two features
FEATURE_N_LAYERS	OpenSURF number of layers
FEATURE_N_OCTAVES	OpenSURF number of octaves
MATCHER_LAMBDA	Game-theoretic matcher payoff lambda parameter
MATCHER_MAX_EPI_DISTANCE	Game-theoretic matcher maximum epipolar distance allowed between two matched features
MATCHER_MAX_ROUNDS	Maximum number of games to be played
MATCHER_MIN_GROUP_SIZE	Minimum cardinality of a winning strategy to be considered valid
MATCHER_POPULATION_THRESHOLD	Winning strategies payoff threshold
NUM_FEATURES_PER_IMAGE	Maximum number of features to extract for each image