in this project will have record bag and measure picture from bag with the RealsenseD435 camera. The scripts are separated into two parts, collector and processor.
the collector is to let the driver collect data effortlessly, with one remote controller.
the processor is creating data from the bag files and the txt recorded, match frames and location, using arcpy API to create shapefile for users.
The further usage is using matplotlib within ArcGIS 10, let users read ROSbag files with hyperlink script function
data collector: will generate folders:bag, log, and live location csv
data processor: from the collected bag and log file, create jpg and shapefile
double click to open the measure exe
Esc/Q :leave
Space: switch between modes
Video mode:
any other key can go to the next picture
Measure mode:
right key to go to the next picture, left will go the the previous picture.
left click, hold to measure, right click to erase
note:go back will take longer to find frames.
the script is mainly based on Realsense, using ROSbag and txt files to record gps location
*the save_single_frame function can't get depth data, show always 0,0,0, therefore I can only use rs.recorder
Hardware: GPS reciever (BU353S4 in my case), Intel D435
the data capture script will automatically generate three folders: bag, foto_log
it will put the recorded bag in bag format,
information of index,color frame number, depth frame number, longtitude, lattitude, day, month, year, timewill be saved in txt format in foto_log
with Intel D435 & GPS reviever
it will take picture every 25 meters while driving
Space: take on photo manually
P: pause auto mode, manual picture still available
R: start a new record
Esc/Q: shutdown
it will create 2 csv files, one for current location and one will show all the locations
The new version: now it will write two csv files, and is used to open in QGIS, set open layer with csv, and the rendering frequency 1-5 seconds depends on the performance of your hardware. The reason is QGIS can render the huge amount of point data better than Google Earth
in QGIS: Add Delimited Text Layer and set render frequency to 1 second for live.csv, foto_location.csv
kml will generate live location as live, and the location with recorded places in Foto.kml use Google Earth: Temporary Places/Add/Networklink link to folder/kml/ live for location Foto.kml for foto point
to prepare the data, use the data_prepare.py
click the three buttons in the order left to right
select the folder with the recorded data(the one contains bag & foto_log)
start with create jpg, it will generate the pictures from the bag file
then the create list will read the jpg files and match a list with the coordinates of the time recorded
with geotag it will read the match list and write the coordinates to the photos.
1.generate jpg using openCV to generate JPEG files, one sharpening filter is added.
kernel = np.array([[-1, -1, -1],
[-1, 9, -1],
[-1, -1, -1]])
*due to disk writing speed, sometimes the realsense api will miss some frames, this script loops through the bag files for two times. can be run multiple times if needed
*this script will skip existed files
2.generate shp because of latency, the frames recorded and the frame number in foto_log can jump in a range around +-2, for example: in log is frame no.100 but it could happen that the recorded is 99-102, so one extra step to match the real frame number to the log file is needed.
after the matcher.txt is generated, this is the complete list of points taken, use MakeXYEventLayer, and FeatureClassToShapefile_conversion will save a shapefile.
*the earlier version has match the frame in a 25ms timestamp for Color and Depth to make sure the frames are matching, the current version removed it, since the latency is acceptable
3.geotag
according to the matcher.txt, photos EXIF will be written, the Lontitude and Latitude
in Arcgis use hyperlink and show script, select python put in
import subprocess
def OpenLink ([jpg_path] ):
comnd = 'measure.exe -p {}'.format([jpg_path])
subprocess.call(comnd)
return
The path of the measure.exe should be the full path
End with an example of getting some data out of the system or using it for a little demo
in here we use subprocess because the current realsense library will freeze after set_real_time(False) and ArcMap don't support multithread nor multicore therefore we can't use the simple import but have to call out command line and then run python
technical: It will base on the matcher.txt and find the corresponding frame number, to get the correct measure, therefore it takes longer
- Che-Yi Hung - Initial work - soarwing52
Creator:
Jay(Che-Yi), Hung
Creation Date: 2019/10/08
Camera: Intel Realsense D435