We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

The use of open source software for monitoring bee diversity in natural systems: the BEEMS project

00:00

Formal Metadata

Title
The use of open source software for monitoring bee diversity in natural systems: the BEEMS project
Title of Series
Number of Parts
351
Author
Contributors
License
CC Attribution 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Identifiers
Publisher
Release Date
Language
Production Year2022

Content Metadata

Subject Area
Genre
Abstract
The use of open source software for monitoring bee diversity in natural systems: the BEEMS project This work wants to highlight the results obtained during the BEEMS (Monitoring Bee Diversity in Natural System) project, which the main goal was to answer the following question: Which biotic and abiotic indicators of floral and nesting resources best reflect the diversity of bee species and community composition in the Israeli natural environment? To this end, the research was oriented towards the cost-effectiveness analysis of new aerial geomatics techniques and classical ground-based methods for collecting the indicators described above, based only on open-source software for data analysis. The study involved the Israeli and Italian teams, focusing the attention on two complementary study systems in central Israel, the Alexander Stream National Park, an area undergoing an ecological restoration project in a sandy ecosystem, and the Judean foothills area, to the South of Tel Aviv. In each study system, different surveys of bees, flowers, nesting substrates and soil, using classical field measurement methods have been conducted. Simultaneously, an integrated aerophotogrammetric survey, acquiring different spectral responses of the land surface by means of Uncrewed Aerial Vehicle (UAV) imaging systems have been performed. The multispectral sensors have provided surface spectral response out of the visible spectrum, while the photogrammetric reconstruction has provided three-dimensional information. Thanks to Artificial Intelligence (AI) algorithms and the richness of the data acquired, a methodology for Land Cover Classification has been developed. The results obtained by ground surveys and advanced geomatics tools have been compared and overlapped. The results are promising and show a good fit between the two approaches, and high performance of the geomatics tools in providing valuable ecological data. The acquisition of the indicators identified in the planning phase took place through several measurement campaigns conducted in the period between February 2020 and April 2020 located in two areas of interest in the Israeli territory. A total of 934 and 543 wild bees were collected in the two systems under study, respectively. From a geomatics point of view, 8 flights were carried out in the Alexander Stream National Park on 24 February 2020, acquiring approximately 65 GB of 8-bit multi-band images in tiff format. In the Judean foothills area, 11 flights were carried out on 26 February 2020, obtaining approximately 77 GB of tiff images. In addition, in order to obtain a correctly geo-referenced 3D model, a total of 54 Ground Control Points (GCPs) were acquired, of which 27 in Alexander Stream National Park and 27 in the Judean foothills, with a multi-frequency, multi-constellation GNSS geodetic receiver in RTK mode. On the basis of the technical requirements necessary to carry out this project, very high-resolution digital maps (orthophotos, digital terrain models - DTM) were produced through the application and optimisation of photogrammetric and structure from motion (SfM) processes performed on data from different imaging sensors (RGB, multispectral), considering only open-source software. Therefore, considering all the previously defined aspects, in order to plan the data acquisition, the research group defined the flight parameters and instruments, both in terms of aircraft and sensors to be installed onboard, necessary to achieve the project objectives. All the digital cartography generated has been defined in the Israeli reference system, i.e. in WGS84 with UTM 36N cartographic projection. The results are shown in Tables 2 and 3 for the Alexander Stream National Park and the Judean foothills, respectively. The production of very high scale digital cartography allowed the extraction of the necessary data for training the proposed Artificial Intelligence model. These data were applied to two different approaches for automatic land cover classification. The first approach was based on unsupervised classification at the pixel level, while the second approach is based on object classification, i.e. vector polygons describing the boundaries of a real object. The algorithms operate differently on these two types of data, in fact in the pixel-based approach they are applied at the level of the single pixel, while in the object-oriented approach they are applied to groups of homogenous pixels for a given feature. The implementation of all the training and validation phases of the proposed models was based on Python programming language using open libraries for data management (shapely, raster) and learning (sk-learn). The segmentation of the input data is fundamental in the approach in order to define the objects to be classified, therefore the Orpheo Toolbox library was applied. The object-oriented approach was applied for the Alexander Stream National Park site while the pixel-based approach was applied on the Judean Foothills area. For pixel-based classification, a clusterization algorithm, KMeans, was used in an unsupervised manner. The KMeans algorithm clusters the data by attempting to separate the samples into n groups of equal variances, minimising a criterion known as within-cluster sum-of-squares. The algorithm was optimised through a trial-and-error procedure that led to the identification of initialisation parameters. For the object-oriented classification, we proceeded to apply automatic segmentation algorithms based on the analysis of multi-band spectral variability. In particular, the algorithm used is the Large-Scale Mean-Shift segmentation algorithm, which produces a clustered image in which the pixels around a target pixel that present similar behaviour from both the spatial and spectral points of view are grouped together. Then, the procedure vectorizes these clusters and the operator associates a label to each of them for the generation of the dataset. After subdividing the data into training and testing elements, the Random Forest algorithm was used for both approaches and proved to be the most effective in performing the assigned task. The classification results were carried out using different validation metrics such as Precision, Recall, F1 score, etc. that will be presented.
Keywords
Open sourceSystem programmingNatural numberIntegrated development environmentAreaTrailCollaborationismProjective planeSpeech synthesisPresentation of a groupSoftwareComputer fontMereologyContext awarenessExploit (computer security)Open sourcePolygonComputer animation
SoftwareSystem programmingOpen sourceTrailGeomaticsDigital signal processingMathematical analysisComputer-generated imageryObject-oriented programmingProcess (computing)Geomatics1 (number)Mathematical analysisPolygonPresentation of a groupComputer fontSoftwareProjective planeResultantProcess (computing)WebsiteUniverse (mathematics)Open sourceMetropolitan area networkSlide ruleBilderkennungComputer animation
Open sourceTrailSystem programmingSoftwarePressure volume diagramNatural numberState of matterConservation lawNumerical taxonomyWage labourDirected setSystem identificationIntegrated development environmentData integrityPrice indexMixed realityClassical physicsIntegrated development environmentINTEGRALGeomaticsTrail1 (number)Observational studyProjective planeContext awarenessPhysical systemSpacetimeGroup actionCollaborationismMereologyReliefMultiplication signSoftware developerComputer animation
TrailOpen sourceSystem programmingSoftwareGreen's functionIntegrated development environmentPC CardProjective planeIntegrated development environmentContext awarenessPairwise comparisonMathematical analysisDifferent (Kate Ryan album)Universe (mathematics)Disk read-and-write headParameter (computer programming)Conservation lawResultantProcedural programmingFocus (optics)Multiplication signComputer animation
SoftwareSystem programmingOpen sourceTrailMetric systemDigital rights managementProcess (computing)Archaeological field surveyIntegrated development environmentGeomaticsInformationMathematical analysisBildsegmentierungSocial classResultantMultiplication signOpen sourceProduct (business)Projective planePhase transitionLatent heatProcess (computing)CountingMappingSoftware testingGeomaticsFrequencyExtension (kinesiology)Level (video gaming)Term (mathematics)Procedural programmingAreaPixelObject-oriented programmingComputer animation
Field (computer science)ForestPhotographic mosaicTrailPC CardOpen sourceSoftwareSystem programmingProjective planeCondition numberMathematical analysisMereologyNetwork topologyComplete metric spaceIntegrated development environmentPatch (Unix)Extension (kinesiology)Type theoryAreaDifferent (Kate Ryan album)Slide ruleRow (database)CASE <Informatik>Raw image formatComputer animation
TrailClique-widthGreen's functionRange (statistics)Matrix (mathematics)Angular resolutionWeightSpectrum (functional analysis)Bit rateFrame problemComputer wormDimensional analysisDuality (mathematics)System programmingSoftwareOpen sourceoutputGame controllerRange (statistics)Point (geometry)Model theoryLevel (video gaming)Spline (mathematics)Computer animation
TrailMultiplicationMusical ensembleComputer-generated imageryPoint (geometry)Error messageControl flowSystem programmingSoftwareOpen sourceDifferent (Kate Ryan album)Spline (mathematics)SoftwareLevel (video gaming)Open sourceTerm (mathematics)AreaPoint (geometry)MappingSystem programmingoutputResultantInformationComputer animation
SoftwareOpen sourceTrailSystem programmingTexture mappingNumberLevel (video gaming)ResultantPatch (Unix)Photographic mosaicSubsetPhase transitionDimensional analysisField (computer science)Computer animation
Sample (statistics)Model theorySocial classGeometryMathematical morphologyOpen sourceTrailSystem programmingSoftwareInheritance (object-oriented programming)Radiology information systemAreaClique-widthLine (geometry)Entropie <Informationstheorie>Moment (mathematics)Inverse elementCross-correlationVarianceSkewnessKurtosisMoment of inertiaArithmetic meanStatisticsRange (statistics)Green's functionPrice indexSystem identificationIntegrated development environmentSlide ruleSocial classMusical ensembleSpeech synthesisProduct (business)Multiplication signPrice indexInformationSupervised learningRange (statistics)Level (video gaming)Object-oriented programmingType theoryComputer animationProgram flowchart
System programmingSoftwareOpen sourceTrailArithmetic meanShift operatorRadiusComputer-generated imageryRange (statistics)PixelAlgorithmPolygonSlide ruleScaling (geometry)MereologyArithmetic meanType theoryShift operatorPolygonSet (mathematics)Range (statistics)Graph coloringRadiusAreaComputer animation
System programmingSoftwareOpen sourceTrailArchaeological field surveyPixelSample (statistics)PixelMathematical analysisResultantObject (grammar)Type theoryComputer animation
Mathematical analysisForestRandom numberPearson product-moment correlation coefficientCross-correlationPixelSoftwareTrailSystem programmingOpen sourceSign (mathematics)PredictionHarmonic analysisPersonal digital assistantPixelResultantObject (grammar)Cross-correlationArithmetic meanPrice indexNumberMatrix (mathematics)Mathematical analysisForestSlide ruleRandomizationVideo GenieCASE <Informatik>Set (mathematics)PredictabilityHarmonic analysisComputer animation
SoftwareOpen sourceTrailSystem programmingObject (grammar)NumberPixelSocial classType theoryPrice indexResultantInformationDifferent (Kate Ryan album)Slide ruleKlassengruppeCASE <Informatik>Parameter (computer programming)Computer animation
System programmingOpen sourceTrailSoftwareAreaMappingResultantCASE <Informatik>Photographic mosaicLevel (video gaming)Computer animation
James Waddell Alexander IIOpen sourceSystem programmingTrailSoftwareCodePixelArtificial intelligenceArchaeological field surveyGeomaticsCharacteristic polynomialPressure volume diagramModel theorySurfaceDigital signalClassical physicsSpeciesDependent and independent variablesSpectrum (functional analysis)InformationService (economics)ResultantCharacteristic polynomialIntegrated development environmentSpeech synthesisNumberParameter (computer programming)Mathematical analysisMultiplication signPrice indexTable (information)Slide ruleType theory2 (number)Computer animation
Open sourceSystem programmingNatural numberComputer animation
Transcript: English(auto-generated)
Okay, thank you very much, Bianca, for your kind introduction. Welcome to my speech. I'm honored to be the first speaker in this session and the first speaker of the academic track in this room, so I'm really excited to present my work that is in collaboration with Vincenzo di Pietra, he's my colleague at Politecnico di Torino.
He did the most part of this work, so I really want to thank him for his precious work he did. during this project. What's the goal of this presentation? It's to show you how it's possible to exploit open-source software to preserve bees' communities
and to analyze the area where bees live. So, in this context, this is the outline of my presentation. First of all, I introduced the Beams project. You probably don't know anything about this because it's quite a small project, but it's important that I introduce this.
Then, I'll move into the analysis of the data that we have considered. So, I'll give you an idea about the geomatics role. I'm a geomatics man because I'm working at the university and I belong to the geomatics lab at Politecnico di Torino. Then, I'll show you the methodology that we have developed and used for reaching the goal.
I'll describe the sites and the data acquisition, the photogrammetric process, the image analysis that we did using open-source software. I'll show you some results about classification and ground cover and we'll get some conclusions and future steps.
In this slide, you can see all guys which have studied this topic. Some of us are from Politecnico di Torino, other ones are Israeli guys because this project that is called Beams is a project founded in 2019
and is a scientific and technological collaboration project between Italy and Israel. It was granted by the Italian Ministry and the Israel Ministry of Science and Technology Beams is the acronym of monitoring bee diversity in natural systems.
So, the idea is to develop novel aerial and classical ground meters to combine them in order to evaluate biotic and abiotic indicators. You know that bees are really, really important for us because without them, it's not possible to survive for many years here
and it's also difficult to analyze them because they are quite small. The other problem is that the typical or classical methods for analyzing bees and their spaces is time consuming. It's so difficult because we cannot track each bee during its fly,
so it's so difficult. And the idea is to identify, to try to develop some techniques useful to perform these approaches in an automatic way. So, the main motivation of these studies are summarized in these slides.
So, the pollination plays a key role in maintaining global human food supply and ecosystem integrity and it's important in this context to conserve and restore the bees' communities and their spaces, the space where they live.
So, the idea is not only to study them but also to identify the cost-effective environmental indicators in order to be able to focus the attention not on the whole world of bees but to sum up which are the most important parts. And for this reason, the research group is not only composed by geomatics guys
but will show you that is a mix between competencies because the goal is, first of all, to acquire these indicators merging traditional techniques and these innovative ones in order to obtain some methodology that can be applied in a quasi-automatic way.
So, the idea is to process this data and reply this study everywhere. As I said before, I'm Paolo D'Above, I'm one of the principal investigators of this project but the most important work was done by Vincenzo di Pietra.
On the right side, there are two main professors from the Hebrew University of Jerusalem the Faculty of Agriculture for the Environment which worked on two different aspects. Professor Yael Mitchell, which is the head of the Department of Chemistry at this university
worked on the soil analysis in order to identify, to characterize the soil moisture, soil parameters and Professor Yael Mandelik, which belongs to the Department of Entomology
studied the conservation, the biodiversity, the ecosystems where bees live and helped us in order to identify and to collect data the unfilled data in order to be able to perform comparisons in this context. Of course, the time is not so long
so I cannot also show you the comparison results but I'll focus the attention on the procedure and the methodology that we did. So, probably you are familiar with the geomatics terms what we did was to follow the whole procedure starting from the acquisition phase to the processing
restitution and the final results, the final maps that can be broadcasted or provided to stakeholders or people which are not so familiar with open source data with GIS, but if you look at this picture you could probably identify that there are lots of yellow spots
these are the maps that we created in order to count flowers so it's quite important because one of the goals was also to identify how many flowers are in this area for example, this was a portion of an image and perform the count of each pixel
if we can reach so high details so the idea was to perform ground cover classification developing some tools which are rapid, accurate of course, otherwise we can't reach the goals and most important, low cost because if we want to reply this methodology everywhere
we can't use so high cost tools we started with a land surveying like surveying using drones and specific cameras I'll show you which camera we have used we obtained these photogrammetric products
classified these maps we have performed image segmentation the object-oriented classifications in order to provide thematic maps this project was a two-year project that was finished a couple of months ago of course, it was quite short as a time period
and the idea is to be able to have a grant extension in order to test the methodology in other areas because this project started at the beginning of 2020 and as you said, there was Covid so one of the biggest problems we encountered was the problem about flying between Italy and Israel
and luckily, we used the last flight that was landed in Tel Aviv before closing frontiers and for this reason, we were able to collect data unfortunately, we weren't able to perform many other flights
because the restrictions didn't permit us to do this and other analyses we selected two different types of environments the first one is the Leksandr Stream National Park that is based, as you can see here in the north part of Tel Aviv Tel Aviv is more or less around here it's a specific ecosystem
composed by coastal sand areas it's quite big because there are many many hectares we selected only these three patches and each patch has an extension of about 10 hectares we performed 8 flights over this area
and we collected 65 GB of raw data that had to be processed on the right side of this slide you can see the other environment that is totally different you can see almond trees this is settled in the Judean Foothill that is in the south part of Tel Aviv
so the environmental conditions are completely different no sandy areas, but shablands and Maki's ecosystems lots of almond trees these are really important because over these trees can be found bees for pollinations so in this area, we selected again
only three patches and each area has an extension that is comparable to the previous one in this case, we did 11 flights and we collected 77 GB of raw data these data were collected using no really low cost instruments
up to now because the idea was to say let's use the typical or traditional tools that we can find on the market and then we perform a downscaling approach from the typical approach to a low cost unfortunately, we didn't have the possibility to perform second flights so for this reason we selected an aircraft that is a DJI
Materice 210 with a slant range camera this is not an RGB camera this is a multispectral camera which we will describe later and we also set some ground control points following the typical photogrammetric approach
to create a georeferenced map, a 3D model and from the input images, we performed the features matching using ground control points we have defined we have georeferenced the maps
following the Israeli reference system and after this rough map, we did the refinement and for each area, we have obtained 82 million points like a map of course, we have also compared open source software with
non-open source in order to verify how different results can be reached in order to provide also these info to their stakeholders and in order to see also what's the difference in terms of accuracy, precision and quality of these maps then we created meshes look at the
number of possible faces obtained, texturization creation of digital terrain model and digital surface model and at the end, an orthomosaic look at the resolution one centimeter of resolution because we suppose that each flower has a dimension of one centimeter
in this way, the idea was to see if we are able to detect flowers and compare this approach with on-site field sampling, so for each patches the Israeli guys were there and count flowers, not only in 10 actors
but only in a subset of these patches and then we have propagated the number of flowers in order to compare these results then, after the typical the traditional photogrammetric approach, we exploited the Orfeo toolbox yesterday was an amazing workshop, unfortunately I
hadn't the possibility to follow that but Bianca did, so it was really great coupling these toolboxes with Sheikit Learn tools we have created the final product so we have performed an object-oriented classification
in these slides you can see the three main steps in order to obtain the map production first of all the data preparation the supervised learning approach and the map production especially in these two last steps Vincenzo did a huge
work and I really want to thank him so first of all because the time is not so long for my speech I want to show you how many classes we have decided to identify according to the Israeli guys so we have decided to distinguish these ten possible classes
one of them, the hoodie plants have been divided in two main subclasses in order to distinguish better the bees environment and thanks to these slant range camera which is as I said is not only an RGB camera but is able also to
use other bands we were able to perform a feature classification, exploiting all this information from other indicators and in this right side of these slides you can see the main bands that can be sorry, the main indicators that can be created
using this type of sensors all fail toolbox implements the large scale mean shift segmentation and we follow this type of approach so first of all we have selected a spatial radius and color range in order to reduce the area, we
label the images in order to be able to perform the training approach then finally we vectorize the data and in this last part of the slide you can see the
steps that we have performed in order to create a set of polygons with associated features so we followed both pixel and object based approach the pixel based is made with ecologists in order to be able to compare these results so we
performed this type of analysis and let me go into these slides in order to summarize the dataset composition and the results so we started with an array, a matrix that was quite big for pixel based and object based we obtained at the
beginning 41 columns means 41 features and 129 columns for the object based there were too many indicators and staring from this dataset we decided to use the Pearson correlation analysis in order to
drop all indicators with a correlation score of 0.9 so starting from this dataset we reduced the number of columns from 41 to 21 and from 129 to 47 ok, after this feature
dropping we did an analysis using the Gini criterion during the random forest training and we validate our results using the three main indicators that you can find in literature, so first of all the precision which compares the true positive over the true
positive plus the false negative and the meaning is like this of all positive predictions, how many are really positive and recall that is quite the opposite of all real positive cases, how many are predicted positive and combining them we reached
the F1 score that is the harmonic mean of precision and recall by definition so it combines these two previous indicators into a single number so here in this slide you can see the accuracy values that we got for the pixel base and the object base as you can see there is a quite
difference because we reached 0.92 for pixel base and 0.79 for the object base there are quite strange results for example if you look at indicator class 6 you can see that
we weren't able to obtain any information about precision, recall and F1 score because the support number was so low, so it was quite difficult to compute these three indicators but for all other parameters we did quite, we obtained quite good results in some cases also
0.98 if you look these for example for this class also using the object base we got really high values but for example for class number 4 the F1 score was quite different also the number of
classes, the type of class played a crucial role at the end we created classifications maps using also the percentage of the area, the covered area and trying to reply the methodology out of the area that we have studied in order to
generalize the results in this case you can see the orthomosaic that we have created and then different thematic maps that have been obtained this is another example of the second type of environment so the Judean Fodils and the Alexander National Stream Park
and in this slide I don't want to stress your numbers but we got very wide tables with lots of numbers it's not the goal of my speech but you can find all the results if you are interested in you can read the papers and you can find all the details
details, sorry let's move to the conclusion because the time is quite over so I don't want to enhance how many parameters we have studied and how important is the bees environment and the bees ecosystems
so we found three focal habitat characteristics indicators and we have tried to define a methodology for land cover classification that is quite automatic that's quite good we replied these methodologies in other areas even
in the same portion of the Israel territory and we got the same results of course which are the next step try to reply this analysis in other places firstly in Israel, then in Italy and then all around the world trying to generalize the methodology and then reduce the cost
of the instruments that we have used that's all so I hope to be on time so thank you very much for your attention