A Dashboard, which shows the current virtual image would be nice
Furthermore commit a4078e1 was wrong. It looks only for the biggest image in the current virtual image. It should take the biggest image over all runs.
a71340ee8b0b68bac341108b5dfa08d718102cb0 added meaningful title attributes to every page for easier tab navigation
497c1a3ab332b5e5853aa90bbab712237a402ff6 Virtual images can now be scrolled through
eb25f19c51ee9677ca156fd8389a090911e3b52b added arrow buttons to simplify the selection of available cutted plants in the current image
eeccf111dfa7cb3c4872ad9c9ab15d91df637667 Asp net core does not permit the download, but the url creation and creation of the zip file works already
630cf62b3cf3a9b1b38f437c3a17cb1532ecf108 static files allow now the folder download to be directly accessed. Furthermore file names can now be sanitized via an extension method
bb8c8bade0409cc31a9207e6dd2ce88f1f620870 refactored reactive values from store into the photostitching feature folder.
It should be used only for that feature
f14d2e7d125c99db25469893b736aa65912cdf19 the default route redirects now to the dashboard page
72b4d5b9f559666b4c8382ae33cf53352e8a6c5c downloading of large tours works now - tested on 7.4 GB.
It takes a while to compress the result, an updating notification would be nice
13d4c29b73b0808b7bb61b06db8e62e3a5f2896c since the compression part takes a while I introduced a way to see the current status of the compressing status.
When the compression is completed, the file can be downloaded as long, as it remains in the download folder.
After 30 minutes the zips are deleted automatically und must be compressed again
3ba543a55fa98eafe71682c7a63ebb24cbea6a31 zip files are now deleted upon server restart. Else disk space is used up by unnecessary zip archives of old photo tours
8d3f07729d6b940f84bd5c3c2f791f597af4c14b plant summary can now be selected. Added methods for easier enum handling
d8ec3dcce9d0eeb495ef3390a49bde30ddf7b537 intervalls between trips are not constant anymore. This means the next trip has to be fetched to calculate the time between those two. From this timespan the temperature values are taken
ca778227ee9e81eacdba4538cd3500fd7a53b91e the length of the array is out of bounce
f79b5aedd5a612cbda31b27a989a95561e1226b3 updated datamodel and implemented how to find virtual images to process for the photo tour summary worker.
Next is to define a proper datamodel for the exported data_trip tsv file to read and write from
ac340f1781a540981024c4e0f61797ae443b8629 added a metadatamodel struct to hold the required fields for import and export of tsv metadata.
Implementation of import and export must still be done though.
1fe3c132f1e3c0eaf4ad725faf7f7dc4af5bf21e Exporting as TSV works now via reflection and automatic finding of the Attributes to export
37f50e8bb7743204ee8c4d35b6774f95cb49a3a4 empty lists are behaving reasonable. Added corresponding test
0238161b901c92584ee204a0d8d59280c86e4add record structs are not working for a parser, because of immutability to set values via reflection.
Converting to classes again
c1db3c83115fd6048fd354aa69741ea0e8ee6173 tsv parsing seems now to be working in both directions
6af8e044e8c8e4778ae550cbea45c149a0881c8d reading from tsv with empty lists works aswell.
Now the implementation of the data extraction can go on
e669614e767f36b837b372e2a76d331bc3d03a90 12 hours are good, if the user has the option to repack the data again - implementation not yet finished
33cb87d1e886ae4c9c50b12fa6170e762b533f3b Zip packages can now be deleted, for example when rapackaging is wished without waiting 12 hours
0e8f50840de63d8c1d39aec5877db3ac3b490ee1 TSV Export should now work. It uses now the correct property names and not the type names anymore.
Furthermore the celcius conversion had to be done from an int kelvin, aka kelvin multiplied by 100.
49054ba94d4f20fae432d276bc441b3c9ca4d661 wrong order of assignment of width and height
8733ebf0c723301d5e4870d4c3dd14b60fb7b9f6 Green parts of an image are now masked and morphological opened to remove small pixel errors.
Now it is known for each pixel, whether it corresponds to a leaf or not. The temperatures for each pixel can be taken from the rawir-image as it has the same size.
The pixelinfo is then grouped by the corresponding image and another class named PhotoSummaryResult executes then the calculation of descriptors for the image.
27591c2c908779f55987e39e0c737a111888eb5e Creating pictures from the extracted pixel positions does not work yet.
Somehow Rgb2HSV works better than Bgr2HSV. The order of the planes in the image should be bgr, because it is read that via via ImRead.
0a02365e5bf4a696ae749a3c9ca6dc51ff8900a3 sub image of segmented plants are now calculated correctly
02443624bb3ad91a1483ad8e800a5319c3d10428 added calculation of different image descriptors.
Convex hull and perimeter are currently not accurate, as there are not necessarly connections between each leaf after segmentation.
This means the convex hulls might overlap each other and the perimeter is calculated for each overlapping leaf set
41cebb398dcb329347fe2993f6b7ed81dee51464 pixel size and interval between photo trips can now be adjusted without restarting of the phototour
07d0d4b2eb066075403c808f6ffb237bccf4b8bf pixel size in mm is now exported in tsv export aswell
2afd55427c285ad99525b7b1fd0dba2eadf822e0 metadatamodel can now be constructed from missing values aswell.
Removed some image descriptors, which are hard to calculate like perimeter and roundness. Proper segmentation of stems is not easy, when the segmentation should be optimized to yield proper ir values
07ca0eb63061685a8e1a6f49afd36f1e4fc0bc73 added more image descriptor calculations for hsl, leaf detection, darkness detection and proper convex hull calculation by combining all points of the found contours.
Leaf counting is missing though
7c059964c324be5f2e8d34fa66d803d95b9c80bb added a basic estimation of the leaf count by eroding the image untill no contours are present anymore.
The maximum amount of contours after applying an elliptic kernel is used as leafcount
c6701a8dae101c3e8be40e1f73905be0cde7cbfc established a baseline of leaf finding, which can be improved upon
0cf7c5bbdb8a3162f2fd3261d1d7511e64f4f2cc better layout of comparison
a83abd2af0a88def4767ed45b6a477df94500482 Added miss summary on top
d0b2568daa19271acc0c5d00363c15dc35a51de8 3x3 cross is an improvement over the previous kernel for structuring element
88960daf90f5d019730ece1cc6860be9177882ed Descriptors are now stored as json in the database
06f49d1cb8fdb81ba31f8a3f45ec1fe2938afde1 pixel size in mm is now correctly stored in the tsv from the database and propagated to the plant summary calculation
b719b656e00bf7a02f97580d41f18644395494f0 the device temperatures must be included in the database summary aswell.
Added some calculations, which do not yet add the results to the database result
870b64bad06273c27bf017caf64d4d604322223c photo tour summary should be stored along the plant descriptors with thermal device information - without redundance
49ff5373861d5a32d5eff993440845c696e3ec40 Fixed tests after change of the tsv model.
Furthermore outdated tsv models should now yield empty results instead of throwing
c72538d9664da3b90abdd2ccaabf0f0a8d4f0485 empty images are no problem during processing, as there are no green parts. Same happens with dark images
467beaca654a7f109dc09883ba79bb19b9e6f2ec the previous ir offset is now available when doing ir fine adjustment.
This helps when new polygons are cut, but the positions have not changed substantially. Furthermore the previous approach to show all existing ir offsets was not useful at all.
dd510d1831cebc86912eb54e55363d3d0e942cce improved performance by not looking up top and left offsets in a list. Taking a dictionary instead.
5a04b18dafce9eceb1030c5118420c3d536fb74a added more logging, old photo summaries are now removed and the check whether an image was already processed is now done with second accuracy.
a30f1e3e5de88a7d4dd53f0981a5a65e62c46b38 Leaf detection does now work correclty via masking segmenting borders of polygons and comparing those borders with the plant masks
b7d256f9609f1c1d0c5e570fea97e92947d3351e removing unnecessary method for plant detection.
Just comparing with the border data is enough
09863c191d82973b83aed5ab8f9df7c29dd79e14 temperature data should be taken from the previous trip to the current one. This makes interpretation of the raw data much easier
17ff9e13826e41b65db6016eadb6da005cb73dba Photo trip summaries can now be recalculated and photo tour id/ photo trip ids are now added to the summaries for internal lookup purposes.
c7c6e9c6d6c90606105fc1e5a2c2e1ac5552be9e added a view to access the json data in an easier and faster way to delete invalid and obsolete image summaries
Current Setup: 0.2187 mm/px