SfM Quantitative Underwater Imaging Device with 5 cameras (SQUID-5) – Field data from periodic surveys of the Florida Keys and other select shallow water environments

IMAGE AND POSITION ACQUISITION The USGS operates the SQUID-5 system to conduct underwater imagery surveys over shallow water environments with optimized endlap/overlap to facilitate SfM processing (Hatcher and others, 2020 and 2023). The SQUID-5 system is comprised of five synced 5-megapixel (MP) machine vision cameras attached to a rigid frame, along with a custom integrated survey-grade Global Navigation Satellite System (GNSS) receiver, which is towed behind a motorized support vessel. The cameras are triggered by an acquisition computer which is time synchronized using a local GNSS regulated Network Time Protocol (NTP) server. An event mark is subsequently recorded and timestamped by the GNSS for each camera trigger signal. The raw dataset is comprised of the raw imagery, raw dual-frequency carrier phase GNSS data and photo capture event time data. Individual image names conform to the following naming convention: CAM##-YYYYMMDDhhmmss_fff-n[nnn].tif, where ## comprise the last 2 digits of the serial number of the camera that captured the image, YYYYMMDD represents the year (YYYY), month (MM) and day (DD) of capture; hhmmss_fff is the time of capture as hours (hh), minutes (mm), seconds(ss) and fractional seconds (fff) in Universal Coordinated Time (UTC) and n[nnn] is a sequence number (image counter) which is reset at the beginning of each survey line. For example, CAM82-20210508215908_001-115.tif was collected by the camera whose serial number ends in 82 on 2021-05-08 at 21:59:08.001 UTC, and it was the 115th shot of that survey line. Images from the camera whose serial number ends in 01 come off the system named CAM1- YYYYMMDDhhmmss_fff-n[nnn].tif. Images from this camera and any future one that ends in 0#, where # is in the range [1-9]), is renamed using the script below by inserting the missing ‘0’ such that it conforms to the naming convention and becomes CAM01-*. #!/usr/bin/bash for i in CAM1-*.tif; do echo "$i" mv "$i" "`echo $i | sed "s/CAM1/CAM01/"`"; done GNSS positions record the location of the phase center of the GNSS antenna at the moment an image capture is initiated (see Positional_Accuracy metadata element for uncertainty discussion). Lever arm distances are used to translate the position from the GNSS antenna reference point to the nodal point of the camera lenses; these are provided relative to the coordinate systems of each camera, where the camera is 0,0, and the axes run as follows: X axis runs from left (negative) to right (positive) Y axis runs from bottom (negative) to top (positive) Z axis runs from front of lens (negative) to back of camera (positive) The SQUID system hardware is oftentimes disassembled for transportation and reassembled at the survey site. It is also updated as new technological advances become available, which may include new cameras and/or new GNSS system components. Although every effort is made to reassemble the system in exactly the same positions, both these events cause changes to the lever arm distances. The methodology by which the original lever arm measurements were determined is described in Hatcher and others (2020). Starting lever arm measurements of the SQUID system are listed below; these distances are then refined during the SfM process. Lever Arm Distances: Center camera: 0.034 m, 0.011 m, and 0.84 m (x, y, z) Forward camera: -0.010 m, -0.594 m, 0.762 m Rear camera: 0.131 m, 0.559 m, 0.754 m Left camera: 0.273 m, -0.109 m, 0.916 m Right camera: -0.311 m, -0.029 m, 0.911 m Estimated uncertainties in the lever arm distances are 5 cm in each direction for the outboard cameras and 2.5 cm in each direction for the nadir (center) camera. Equipment specifications of the different SQUID versions and the dates they were in use are listed below. VERSION 1.1 Survey Dates Valid: 2019-07 to 2019-12 Camera Model: Teledyne FLIR BlackFly BFS-PGE-50S5C-C Lens focal length: 6 mm GNSS Receiver: Trimble R7 GNSS Antenna: Trimble Zephyr 2 VERSION 1.2 Survey Dates Valid: 2020-01 to 2022-06 Camera Model: Teledyne FLIR BlackFly BFS-PGE-50S5C-C Lens focal lengths: Center cam 8 mm, outboard cams 6 mm GNSS Receiver: Trimble R7 GNSS Antenna: Trimble Zephyr 2 VERSION 2 Survey Dates Valid: 2022-07 to Present Camera Model: Teledyne FLIR BlackFly BFS-PGE-50S5C-C Lens focal lengths: Center cam 8 mm, outboard cams 6 mm GNSS Receiver: Spectra Geospatial SP90M GNSS Antenna: Spectra Geospatial SPP 135000.00

IMAGE AND GNSS QA/QC Quality control of the raw imagery and raw GNSS data occurs in the field where the entire processing workflow is performed on the day of collection. A preliminary processing of the GNSS data is done against either the closest continuously operating reference station (CORS) or a locally-established base station, at 1 hertz (Hz), using either broadcast or ultra-rapid ephemerides and assessed by reviewing plots and quality metrics from NovAtel’s GrafNav software. The resulting image positions are then ingested into Agisoft Metashape along with the raw imagery where a medium quality point cloud and medium resolution DEM are produced. If any issues that preclude successful model building are detected during this field processing, the batch is discarded, and the data is re-collected. If an issue is found by the data processor that affects two or fewer cameras, a re-collect is not performed, but the images from those cameras will be discarded during final processing. Hatcher and others (2023) determined that the SQUID-5 has enough redundancy to build high quality high-resolution models with as few as 3 cameras operating. Data from the local base station is converted to Receiver Independent Exchange Format (RINEX) using a proprietary utility provided by the manufacturer of the GNSS receiver, down-sampled to a 10 or 30-second interval using TEQC (https://www.unavco.org/software/data-processing/teqc/teqc.html) if necessary to comply with file size limits, and submitted to the National Oceanic and Atmospheric Administration (NOAA) Online Positioning User Service (OPUS) website (https://geodesy.noaa.gov/OPUS/) for validation. This website computes the position of the uploaded data and ties it to the high-accuracy National Spatial Reference System (NSRS) coordinate system.

KMZ CREATION In order to provide a visual map of the survey extent, image capture locations, to aid a user in determining which images they might want to download, and to aid in completing the geospatial information in the collection-level metadata, a Keyhole Markup Language Zipped file (.kmz) is created for each collection using Global Mapper to convert the image positions file to a kmz.

GNSS FINAL DATA PROCESSING Raw GNSS data received by the antenna mounted atop the SQUID-5 are recorded at 10 Hz by a dual-frequency survey-grade GNSS receiver. This reprocessing of the GNSS data uses precise ephemerides (obtained through the GrafNav software from external sources), multiple base stations if possible, and is run at 10 Hz, to achieve higher accuracy image positions than was obtained during the field processing described in the ‘IMAGE AND GNSS QA/QC’ step above. If the collection-level metadata describes the dataset as PROVISIONAL, it means the image positions were computed using a lower accuracy ephemeris. This is sometimes necessary if the imagery is collected to rapidly assess damage from natural or unnatural (e.g. boat groundings) disasters. The trajectories, in combination with precisely recorded image capture event times, are used to generate GNSS antenna positions at the moment each image capture is initiated and is provided for each collection. Images with missing positions are reconciled through direct interpolation based on the image and GNSS time stamps. The positions in the image positions file represent the position of the SQUID-5 GNSS antenna, not the position of the camera or features photographed. To determine actual camera positions, photogrammetric software such as Agisoft Metashape can be used to apply the lever arm offsets (in the camera frame of reference) from the GNSS antenna reference point to the camera lenses.

IMAGERY HEADERS Georeferencing, copyright, and other relevant information is added to the imagery headers of each image using Phil Harvey’s ExifTool. To extract the information from the image headers using ExifTool, the following command can be used: exiftool -n -csv *.tif > allheaders.csv The -csv flag writes the information out in a comma-delimited format. The -n option formats the latitude and longitude as signed decimal degrees.