Automatic Forest Assessment

L A N D C O N S U L T . D EDr. Markus WeidenbachÖhinghaltweg 3D 77815 BühlTel.: +49-7223-9158030Mobil: +49-170-8988393Fax: +49-321-21252410 E.mail: office@landconsult.euInternet: http://landconsult.deInternational consortium of geo-scientists, land use planners, forest engineers and GIS/RS expertsDatum: 10.09.2013Find and Delineate Tree Crowns with e.CognitionRuleset Version 2, description and usageRuleset coyprights at: landConsult.deÖhinghaltweg 3D-77815 Bühloffice@landConsult.euSept. 2013Ruleset DescriptionThis ruleset has been developed and applied in the course of different projects with a big German State Forest Administration and other private forest owners. It is designed to create a polygon shapefile representing single tree crowns. It includes an individual setting function to process all kinds of forest stands including very young and mature stands at the same time. During the processing the stand is divided into homogeneous groups basedon the height, afterwards each groups is being processed separately with different sets of variables. These variables can be adjusted by the user.Following pre-conditions must be fulfilled to deliver the expected results:1. Good quality of nDSM raster layer. The nDSM used during the ruleset development wasprocessed with median filter (3x3), its resolution was equal to 0.5m. It can be processedfrom LiDAR or Aerial Stereomodels. The ruleset is designed to work with rasters ofdifferent pixel size, however the variables have to be adjusted if you choos another pixelsize than 0.5 m. 2. The model should not include any non-vegetation, elevated objects like buildings or polesetc. Such objects, if present within the processing area, need to be filtered out before orafter the processing as they will be most probably treated as trees.If you wish to adjust and further develop the ruleset to meet specific conditions, such as the usage of different RS data or local stand conditions, or if you wish to automatically measure all numeric crown parameters with ATLAS, a helpful tool to analyse the point cloud of the detected crowns, please address your request to office@landconsult.eu. We are pleased to extent and optimize the performance of the entire process to meet your specific needs.1/4landConsult.de: e.Cognition Ruleset to Detect Individual Tree CrownsTo run the ruleset, please do the following steps:1. Run e.Cognition Developer 8.7, and choose “RuleSet Mode”2. Create a new project and add one raster layer representing the nDSM. It is necessaryto set an alias to this layer – "ndsm". Save the project on the hard drive beforeprocessing.3. [Optionally a polygon layer can be added to the project that specifies the exactborders of the processing area. It is necessary to set an alias to this layer –“plot_border”.]4. Import the ruleset file by right clicking on the "process tree" panel and choosing “LoadRuleset”5. [Optionaly modify the parameters]6. Press F5 to start the processing.7. The results of the processing will be stored in the project folder (/results/). Two fileswill be created, one storing the treetops, the second with the crown polygons. Figure 1: Menu to set the variablesThe process of determining the proper crown shape is based on three major stages. First the forest groups are defined, based on their height. For each group, depending on the height, different parameters (variable values) for the next steps.Next the treetops are found as the local maximums on the nDSM raster layer. For each of the stand height group a threshold is defined (variable crowns_search, that has 4 subtypesfor 4 different stand height categories - > 20m, >10m, >5m and <5m) that specifies the pixel distance between maximums. There are some additional filtering steps to improve theresults.After that the treetops are growing several times to create the crown border around each of the treetop. The variable crown_width (_20, _10, _5, _0) determines the maximum width of a crown in pixels in each of the stand height groups (> 20m, >10m, >5m and <5m). The most important is the growing part, which has to be done in a sort of “intelligent” manner. During the growing step, each surrounding pixel around the treetop (or previouslygrown crown as the process is repeated) is analysed and its value (height) is compared to the current values of border pixels within each treetop/crown. If the pixel fulfils the growing criteria (height difference below specified threshold), this pixels becomes a crown part. The process is repeated number of times in a loop until all “candidate” pixels of each found treetop are tested.2/4landConsult.de: e.Cognition Ruleset to Detect Individual Tree CrownsFigure 2: A nDSM with 50 cm resolution, processed from First/Lastpulse Laserdata with a point density of less than 2 pts/dqm and the shape file with the detetected crowns generated by the e.cognition ruleset. Instead of a Laser nDSM you can also use photogrammetrically generated nDSMs.Figure 3: Detected tree crowns resulting from the e.cognition ruleset and and their numeric parameters processed with ATLAS. For each tree crown, ATLAS calculates 16 different numeric parameters, such as the crown and the crown volume plus 8 crown radii.3/4landConsult.de: e.Cognition Ruleset to Detect Individual Tree CrownsReferences:(all References are available Online at: http://landconsult.de/segmentation/download/)Herleitung von Kronenparametern für Einzelbäume anhand von ALS, TLS und Luftbildstereomodellen am Beispiel standortskundlicher Versuchsflächen in Sachsen Vortrag anlässlich des Workshop "Multisensorielle Methoden zur Rekonstruktion der dreidimensionalen Waldstruktur", TU Dresden, 04. Februar 2013. Erfassung von Einzelbaumparametern mit Airborne-Laser-Scanning-Daten. Publikation in AFZ - Der Wald, 21/2012, S. 12 - 15. Machbarkeitsstudie zur photogrammetrischen Erzeugung von hochwertigen Waldoberflächenmodellen mit konventionellen digitalen Luftbildern.Abschlussbericht vom 12.10.2011 über das vom Wirtschaftsministerium Baden-Württemberg geförderte Innovationsprojekt 20114/4landConsult.de: e.Cognition Ruleset to Detect Individual Tree Crowns