Stereo model selection

SURE automatically analyzes the block configuration in order to find suitable stereo models to match. The most important settings are the following ([default setting]):


1) Stereo model filtering based on the exterior orientation (controlInvalidate)

a. Principal viewing ray of each camera (AnglePrincipalTest)

i. Filtering min/max angle between cameras [on]

b. Base line between cameras (distance) (CamDistTest)

i. Filtering min/max baseline [off]

c. Value in user connectivity matrix (ThreshConMatTest)

i. Filtering min/max value [off]

 

2) Stereo model selection in respect to MaxModels / MinModels (controlInvalidate)

According to the maximum / minimum model definition, stereo models are selected by analyzing the following values:

a. Base line between cameras

i. Closest cameras [on] (ReduceModelsCamDistNearest)

ii. Equal distribution of baselines [off] (ReduceModelsCamDistEqual)

b. Value in user connectivity matrix

i. Smallest values [off] (ReduceModelsThreshNearest)

ii. Equal distribution of values [off] (ReduceModelsThreshEqual)

 

3) Stereo model selection based on data initialization on low resolution (controlInit)


Within the Initializer module, the dataset is processed on very low resolution to retrieve information of the surface, which

allows further analyzing. Rectification, matching and point generation can be activated. This module is not activated by default,

but for all AERIAL scenarios. Subsequently, stereo models can be filtered – e.g. by applying a threshold to a minimum successfully

matched overlap (MinPercMatched or -minperc).

 

Manual stereo model selection


Stereo models can also be defined manually by a user connectivity matrix. The connectivity matrix encodes the relation between images and can by passed as a

text file using the flag “-con”. An example connectivity matrix can be found in the List.txt file, which is generated automatically by default. In such a connectivity matrix,

each row and each column are related to an image in the dataset – e.g. the first row of the matrix represents the first image in the image list. The relation between two

images is thus stored in each element of the matrix – e.g. the value in row 1, column 2 depicts the relation between image 1 and image 2. Thus, each connectivity

matrix is squared and symmetrical. Values greater than 0 indicate a connection and will lead to a stereo model. You can deactivate all filters on the matrix using –confix.