Quality Check is a feature that control the uploaded data from a Data Capture task. It has been implemented to assure that the taken pictures matches what have been planned on the Data Capture task and it also enables to have a high quality input dataset before the analytics.
Quality check is a feature activated at company level with options to activate/deactivate particular checks that would not be relevant and always fail.
1. Module Accessory
If the company has this feature activated, then during the upload on a Data Capture task, the Quality Check step will appear as in the third picture:
|Step 1||Step 2||Step 3|
Here are the different parameters that are verified during the Quality Check :
- a - Check if the Date and time acquisition match the scheduled date on the Data capture task
- b - Check the acquisition envelope (the projection of the pictures on the ground should cover the area of interest)
- c - Check if the used sensor matches with the planned one (in the task)
- d - Check if the altitude matches the altitude planned for the task
- e - Check the flight speed matches the speed requested for the task
- f - Check that the Pictures orientation angels (pitch, roll) are within acceptable ranges
- g - Check the local time to have a proper sunlight angle
- h - Check if there are missing pictures (missing data from one of multiple sensors at a particular point, or hole in the acquired trajectory)
Yet, the available checks are the a, f, g, h.
a. Date and time schedule
The module checks if the picture time fall inside the scheduled time window of the Data Capture task. If the time and dates do not match, a warning will be shown to the user.
b. Acquisition envelope
An envelope of all image projections on the ground is computed and a buffer is added to this enveloppe : the buffer size is half the theoretical size of an image on the ground (using the task GSD and sensor width or using sensor pixel size and height above ground ).
The planned area on the task should be contained in the envelope with its buffer.
c. Sensor check
In this check, the sensor model is extracted from the images and the module finds the corresponding sensor in a reference Data base.
This check will fail if the sensor type cannot be found or does not match with the available sensors in the data base.
If bands are defined in the Data capture task, the module compares the bands for the sensor in the reference database to the required bands and issues a warning if the bands do not match. This check also fails if the bands for the sensor used in the acquisition cannot be found in the reference database.
d. Altitude check
If the height above ground is present in the images, the quality check module will compute a robust mean (e.g median) of values and compare it to the requested height on the Data Capture task.
The relative error should be below 15% otherwise the check will fail.
e. Flight speed check
The module is estimating the flight speed using images positions and the time. This estimated flight is compared with the planned flight speed.
f. Picture Orientation check
An histogram of roll and pitch values are built by the module.
For all pictures the roll and pitch values are checked and should fall into the following bins:
|Angles||- Infinity||- 25 °||- 15 °||15 °||+ 25 °||+ Infinity °|
g. Local time/sunlight angle check
The relative position of the sun in the area of acquisition, at the time of the start and end of the data collection is computed.
The test fails if part of the flight is done in low light conditions.
The test issues a warning if part of the flight is done during the “hotspot” hours (when the sun is very high in the sky).
Note that this computation also depends on the sensor’s Field of View angle.
h. Missing pictures
This test analyses the position of all the pictures available after upload, identifies the flight axes, extracts the sensor used for the acquisition.
In a multi-sensor/multispectral acquisition it checks that images in all the bands are available for all trigger events.
It checks for holes in acquisition trajectories by computing the average flight time between pictures along each flight axis.
In both cases any missing picture detected triggers a warning, and the approximate positions of missing images are reported.