implementing additional lidar sensors

[forkozi]

Implementing a sensor model in cBLUE consists (1) defining a laser geolocation equation and (2) specifying supplemental sensor-specific parameters. Currently, cBLUE supports one, generic sensor model, which was developed for use with the Riegl VQ-880-G lidar sensor. With slight modifications, the generic sensor model can be applied to other lidar sensors. Custom sensor models (i.e., those employing different laser geolocation equations), could also be created and integrated within cBLUE.

Using cBLUE’s Generic Sensor Model
The bulk of cBLUE’s generic sensor model is defined in the SensorModel class, which defines the laser geolocation equation and associated polynomial-surface error modeling. Assuming the same laser geolocation equation and associated polynomial-surface error modeling are to be used, the SensorModel class doesn’t require changes for it to be used with another lidar sensor. The following supporting parameters, however, do need to be edited.

• Sensor-Model Parameter Uncertainties
The a priori uncertainties (standard deviations) of the parameters in the sensor model laser geolocation equation (a, b, and rho) are currently specified in the Merge class.
https://github.com/noaa-rsd/cBLUE.github.io/blob/b1eb18315113d62f7df829deec78230faa690a04/Merge.py#L44-L46

• Sub-aqueous TPU Look-up Tables (LUTs)
The sensor-specific lookup table used in computing the subaqueous component of the total propagated uncertainty is currently specified in the ControllerPanel class.
https://github.com/noaa-rsd/cBLUE.github.io/blob/b1eb18315113d62f7df829deec78230faa690a04/CBlueApp.py#L515

One refactoring strategy for generalizing the current generic sensor model implementation is to consolidate the supporting parameters within the SensorModel class and instantiate SensorModel objects with sensor-specific information read from a json file, per the GUI dropdown menu in the CBlueApp class.

Using a Customized Sensor Model
The existing cBLUE SensorModel class implements the sensor model as described by Eren, F. et al (2019), which parameterizes the scan pattern using two angles, a and b, and includes polynomial-surface error modeling. Incorporating other sensor models (i.e., laser geolocation equations) into cBLUE would require some sort of refactoring of the SensorModel class and the associated Jacobian class, which provides the functionality to calculate and evaluate the Jacobian (i.e., matrix of partial derivatives) required to calculate TPU.