RADAR

RADAR scheme in Flysafe database

Data model for Motion Analyses images collected with ROBIN-system

Originally Hans van Gasteren, May 2007

Introduction

Radar data are one of the most important measurements in Bird Avoidance System (BAS) and Fly Safe project. Until now data are archived in structured files instead of in a database. In this document data model for Motion Analysis images of the ROBIN4 system is designed, based on file structure of ROBIN4-MA images as defined in High Level Design Document of ROBIN4 (HLDD, TNO 2006). MA (Motion Analysis) images are summated images of ten radar antenna rotations and are recorded twice per hour, two radar beams per radar, 3 radars in total (equalling 12 images per hour). The data comprise all recorded radar echoes with summated intensities, plus rain clutter masks, land clutter masks and all separate objects of all recognised tracks. The huge amount of data per image (360°, 150km range, 10Mb) require a flexible database. Numbers of tracks can be as high as 15.000 tracks per image, with a mean about 1000 tracks. Per day we collect 48 images * 2 beams * 10Mb * 3 radars equipped with Robin  3Gb data. Because the database must be readily and speedily accessible, only aggregated data can be recorded in the database. Original data will have to be recorded in a file system. The level of aggregation is still under debate, as the database should also allow a certain level of reanalysis. It is not desirable to store all original recorded echoes and raw radar data, which are the bulk of the data, in the database. Because the number of tracks is rather low ( 1000) for most of the images, it may be sufficient to store summaries of each track. This will lead to an estimated data reduction of 90%. In the near future, already during precursor phase of FlySafe project, bird echo tracks will be recorded continuously. This means that if we want to store individual tracks, also positions of each antenna rotation is stored and database could be increasing much faster than it does now.

Data model for MA images collected by ROBIN4 system .

RADAR

The radar table contains details for the different radars.

RADAR_ID	unique radar id
RADAR_NAME	logical name
LATITUDE	degrees
LONGITUDE	degrees
X_POSITION	in meters, rijksdriehoekmeting
Y_POSITION	in meters, rijksdriehoekmeting
Z_POSITION	In meters, rijksdriehoekmeting
ALTITUDE_ANTENNA	In meters with respect to sea level
Radar type:
MIN_RANGE	In meters
MAX_RANGE	In meters
RANGE_RESOLUTION	In meters, normal MPR 30m
AZIMUTH_RESOLUTION	In radials azimuth if horizontal radar else elevation
REVOLUTION_TIME	In seconds, normal MPR 10s
Transmitter:
MEAN_CARRIER_FREQUENCY	In GigaHz
PEAK_POWER	In dB
PULSE_LENGTH	In μS
PULSE_REPETITION_FREQUENCY	In Hz
Antenna:
TYPE	circ, rectangular, omni
VERTICAL_ILLUMINATION	uniform, parabolic, parabolic², cosec²
POLARISATION	(horizontal, vertical, circular)
TRANSMITTER_GAIN	In dB

COVERAGE_DIAGRAM

(for each azimuth horizontal elevation is given).

COVDIAG_ID	Reference to RADAR_ID of table RADAR
AZIMUTH	Radians (0-2PI)
ELEVATION	Radians

BEAM

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID
BEAM_NAME
BEAM_ELEVATION	Elevation of beam (radians)
EL_BEAMWIDTH	Beam width (radians) in elevation
AZ_BEAMWIDTH	Beam width (radians) in azimuth

IMAGE_REQUEST

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID	Reference to BEAM_ID of table BEAM
IMAGE_REQUEST_ID
WINDOW_MIN_RANGE	Measurement range (m) of window
WINDOW_MAX_RANGE	meters
WINDOW_MIN_AZIMUTH	radials
WINDOW_MAX_AZIMUTH	radials
Motion_analysis_parameters	Note: for CMA different
RAIN_CLUTTER_MODE	Automatic, off, on
LAND_CLUTTER_MODE	Automatic, off, on
DETECTION_METHOD	Peak detection, threshold
MIN_TRACK_MEMBERS	2 ..10
ALLOWED_INTERVAL	0..9
MIN_MEMBER_SIZE	1..?
MIN_SPEED	5..50
MAX_SPEED	5..50
ALLOWED_SPEED_DEFLECTION	5..50
TANGENTIA_SPEED_DEVIATION	… part of score function
RADIAL_SPEED_DEVIATION	… part of score function
MASS_DEVIATION	… part of score function
BONUS_SCORE_PROBABILITY	… part of score function
PATH_FRACTION_THRESHOLD	… part of score function
RELAXATION_FACTOR	0.0..1.0 … part of score function
RANGE_RESOLUTION	Correction with respect to default range resolution
AZIMUTH_RESOLUTION	Correction with respect to default azimuth resolution

SUBWINDOW

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID	Reference to BEAM_ID of table BEAM
IMAGE_REQUEST_ID	Reference to IMAGE_REQUEST_ID of table IMAGE_REQUEST
SUBWINDOW_ID
SUBWINDOW_MIN_RANGE	meters
SUBWINDOW_MAX_RANGE	meters
SUBWINDOW_MIN_AZIMUTH	radials
SUBWINDOW_MAX_AZIMUTH	radials

IMAGE_RESULT

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID	Reference to BEAM_ID of table BEAM
IMAGE_REQUEST_ID	Reference to IMAGE_REQUEST_ID of table IMAGE_REQUEST
IMAGE_RESULT_ID
ACQUISITION_TIME	Date and time of request
ERROR_STATUS	Status given back by Robin-system
CORRECTION_LEVEL	False alarm rate, FAR correction factor, important value, should have own graphical representation through out the year and correlate with the weather conditions
NR_OF_TRACKS	Number of tracks in image
RAIN_MASK	Rain clutter mask. Area were bird detection is switched off
LAND_MASK	Land clutter mask. No bird detection has been achieved in this area
IMAGE	Jpeg image of data (high resolution ) or georef TIFF

SUBWINDOW_RESULT

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID	Reference to BEAM_ID of table BEAM
IMAGE_REQUEST_ID	Reference to IMAGE_REQUEST_ID of table IMAGE_REQUEST
IMAGE_RESULT_ID	Reference to IMAGE_RESULT_ID of table IMAGE_RESULT
SUBWINDOW_ID	Reference to SUBWINDOW_ID of table SUBWINDOW
AREA	km²
LAND_CL_PERC	% land clutter in image
RAIN_CL_PERC	% rain clutter in image
CLUTTER_PERC	% land or rain clutter in image
TOTAL_MASS_DENSITY	Total mass density per km², before MA
LAND_MASS_DENSITY	Mass density per km², after MA in land clutter mask
RAIN_MASS_DENSITY	Mass density per km², after MA in rain clutter mask
CLUTTER_MASS_DENSITY	Mass density per km², after MA in rain or land clutter mask
BIRD_MASS_DENSITY	Mass density per km² of bird tracks
BIRD_ECHO_DENSITY	Mean bird echo density per km²
BIRD_MEAN_DIRECTION	Mean direction of all bird tracks in sub-window radians
BIRD_MEAN_SPEED	Mean speed of all bird tracks in sub-window m/s
Other params wrt quality assessment	For each sub-window, filled in during second pass or as software module (which can be modified) each time data is retrieved.

TRACK_RESULT

The TRACK_RESULT table contains tracks of birds that have been detected by the ROBIN4 system. It also contains as arrays all the different points in a track data. That data used to be in the TRACK_OBJECT table. The TRACK_RESULT table is a master table and the real data is contained in the inherited tables TRACK_RESULT<YEAR><MONTH>. Otherwise the TRACK_RESULT table would become too big.

RADAR_ID	Reference to RADAR_ID of table RADAR
BEAM_ID	Reference to BEAM_ID of table BEAM
IMAGE_REQUEST_ID	Reference to IMAGE_REQUEST_ID of table IMAGE_REQUEST
IMAGE_RESULT_ID	Reference to IMAGE_RESULT_ID of table IMAGE_RESULT
TRACK_RESULT_ID
DATE_TIME	Starttime (date + time) of track (for MA equal to ACQUISITION TIME, for CMA different)
TRACK_RANGE	Start position of track (m)
TRACK_AZIMUTH	Start position of track (radians)
TRACK_MASS	Mean mass of bird track
TRACK_SPEED	Mean speed of bird track (m/s)
TRACK_DIRECTION	Mean direction of bird track (radians)
TRACK_OBJECTS	Number of objects in track (0 – 10)
TRACK_SOURCE	MA / CAM
OBJECT_LATITUDE	Array of latitude position of object
OBJECT_LONGITUDE	Array of longitude position of object
OBJECT_MASS	Array of mass (reflection) of object
OBJECT_SIZE	Array of size of object
OBJECT_RHO	Array of rho coordinate of object
OBJECT_PHI	Array of phi coordinate of object

Quality parameters

Deze gelden voor elk subwindow opnieuw. Dat betekent inderdaad meerdere keren per beeld.

1. Correction level wrt FAR. Dynamische range bepalen van een subwindow en vervolgens een kengetal tussen 0-100 of 0-1 uitrekenen met betrekking tot kwaliteit. Deze waarde geldt voor een heel beeld en niet voor elk subwindow
2. Percentage landclutter in subwindow. Ook dit hoort constant te zijn, wanneer dit buiten marges valt zal dit de kwaliteit omlaag brengen. Via analyse bepalen wat relatie is met 100%.
3. Te veel vogels. Dit valt af te leiden uit de samenvattende gegevens, volgens:
   1. LAND_MASS_DENSITY is ongeveer normaal, dan moet ook
   2. TOTAL_MASS_DENSITY - LAND_MASS_DENSITY ongeveer gelijk zijn aan BIRD_MASS_DENSITY. Is dit niet het geval (buiten marges), dan te veel vogels en kan nieuwe dichtheid en massa alsvolgt worden berekend:
      1. BIRD_MASS_DENSITY = TOTAL_MASS_DENSITY - LAND_MASS_DENSITY
      2. BIRD_ECHO_DENSITY kun je op twee manieren uitrekenen. Wanneer deze waarde 0 is moet dat via een van te voren bepaalde regressielijn, anders kun je dit uit de data zelf berekenen volgens TOTAL_MASS_DENSITY - LAND_MASS_DENSITY * (BIRD_ECHO_DENSITY / BIRD_MASS_DENSITY).
   3. Uiteraard moet de dichtheid en massa met een lagere kwaliteit in de database worden gemerkt.
4. Dikke vogelecho’s worden als regen gezien. Dit valt af te leiden uit het regenfilter: (1) veel losse stukjes beeld en (2) oppervlak van deze stukjes is klein. Daarnaast moet de gemiddelde massa per echo ook groot zijn: BIRD_MASS_DENSITY/ BIRD_ECHO_DENSITY >> (Wier >700). In dat geval bereken je de nieuwe massa en vogeldichtheid volgens 3.2.1, respectievelijk 3.2.2 (versie 2). Nieuwe getallen moeten met een lagere kwaliteit worden gemerkt.
5. Foute tracks aan de rand van regenbuien. Het gemiddeld aantal objecten in de tracks zal laag zijn.
6. Aantal tracks wat bij laatste iteratieslag hoort in relatie tot totaal.

De vraag is of we deze berekeningen ook opslaan in de database, of dat de regels als software module bij het ophalen van de data worden meegegeven? En dan tot slot een correctie op detectie in afstand en zijaanzicht en de daarmee veranderende hoogte. Dit kan echter als een aparte software module in het hoogtemodel worden opgenomen en hoeft de punten 1 t/m 6 niet in de weg te staan.