General Adjustment controlling

This section controls the actual adjustment. Use Run Adjustment to perform a bundle adjustment with the selected images using their current active sensor model and the selected control point array. Only the selected image point observations and 3D points are considered.

Adjustment control panel

After an adjustment has been performed, some information are displayed under Results.

s0 describes the weight of an observation after the adjustment process with an original weight of one before the adjustment has been performed. This value should generally be around 1.0, meaning that the stochastic and functional model is correct and no gross errors are present. In other words, your observations have been introduced with good a priori accuracy information and a set of correct parameters has been selected.

If s0 is significantly different from 1.0, review the accuracy information used for the image point observations and the 3D control points. A value lower than 1.0 indicates that observations have been introduced too pessimistically. When a value higher then 1.0 occurs, loosen the introduced accuracy of the observations. A second reason for a s0 significantly different from 1.0 could be wrong parameters. Gross errors will also lead to a s0 different from 1.0. Use the robust estimation to locate gross errors. See robust estimation further down.

n observations gives the number of observations considered in the adjustment process. An image point introduces 2 observations, whereas a control point 3 observations, and an orbit point introduces 6 observations.

n parameters describes the number of parameters determined during the adjustment process. The redundancy considers the observations, parameter constraints and direct observations handled in the adjustment process.

When a robust estimation has been performed, gross errors will show the number of gross errors found and eliminated during the iteration process.

Detailed information about the adjustment can be found in the protocol file. Click on View Protocol to open a protocol viewer. See Adjustment Protocol for more information.

Robust Estimation

Robust estimation is a technique to locate gross errors within the observations. For this purpose the quotient of residual v and sigma sv of the observation is computed and statistically analysed. If a quotient is above a certain threshold the observation is disabled for the next iteration. That process is continued until every quotient for all observations is under the threshold or too many observations have been eliminated. Before using this technique review the following preconditions:

• The gross errors are not too large that the whole adjustment fails converging. Run the adjustment without robust estimation enabled to ensure convergence.
• The robust estimation uses the stochastic model & the a priori accuracy information of the observations to decide whether an observation is to be treated as a gross error. If required review the accuracy information of your observations. Too optimistic accuracy settings will lead to more observations marked as gross errors.
• Only observations with enough redundancy can reliably be detected as a gross error!

Only use the robust estimation when the preconditions are fulfilled!

There are two parameters used with the robust estimation. max gross errors % sets a threshold for the maximum number of observations to be eliminated during the iteration process as a percentage. The process stops if more observations are marked as gross errors. Gross error threshold determines the quotient v/sv for which an observation is treated as a gross error. The default value is 4.0. There is generally no need to change this value.

Image point and control point observations which are detected as gross errors will be highlighted red in the corresponding tables. For every further adjustment these observations will be excluded, even though robust estimation has been disabled. Use Reset robust to include these observation again.

Forward Intersection

Forward Intersection calculates 3D points from existing sensor model parameter. The sensor model parameters remain unchanged.

Back to Menu Adjustment