Hi,

At the first step, it depends on the type of satellite image you want to work with, As they have different abilities and spatial resolution. If your image have not the ability to distinguish individual canopies, I would advice to use experimental indirect relationships (may be allometric relationships) between a variable like leaf area index (LAI) and crown diameter (or a variable that you think would have a robust relationship with crown diameter) and then estimate crown diameter via satellite images indirectly using, for example, NDVI- LAI relationship.

Regards,

Omid,

Hi,

If you have a Very High Satellite resolution images, I suggest use to use Object-Oriented Classification techniques.

Good luck,

Thanks for your answers :D.

The NVDI - Variable relationship sounds very interesting!

And yes, we have access to some VHS images. We will look Object-Oriented Classification techniques :D.

We have used with different approaches to do this.

Check this paper; Article Using aerial photography to estimate wood suitable for charc...

We have "developed" specific models that link tree-crown surface with other dendrometrical variables.

If you need any more help, please contact me

Hope it helps!

AGT

Alberto Gomez-Tagle Thanks for the paper Alberto. ! Regarding your approach, we have several limitations with our study in relation with the images and the field information.

Surprisingly (Well, taking into account the type of mayors that Bogotá and other cities in Colombia has had, it's not a surprise), the inventory of several cities in Colombia only have trees height and their spatial position (I don't know why). We have access to some WorldView 2 and 3 images of some cities, but we can't afford (We don't have the money) to do field surveys.

We wanted to search methods that need minimal or null field information to calculate variables like biomass or leaf area of individual trees regardless their species (Or may be using physiological or ecological information of them), and it's clear that the satellital images can be a great tool for this.

Other thing is that looking ( In recomendation of Reda Yaagoubi ) for Object-Oriented Classification techniques, several works have created algorithms to automatically delineate tree crowns in images, but most of these algorithms (Machine Learning, and OBVIA) are in template zones and in forest ecosystems. In cities, that procedure seems harder and although the topic it's not new, is a bit hard for us (Forestry engineering undergraduates) to understand and develop, but we are trying our best!

We will search for more about allometric ecuations, the using of NDVI, and about the OBVIA approach :D

Thanks for your answers!.

Daniel Fernando Florez, Altough we used visual interpretation and manual delineation of tree crowns for the final models in our study, we tested several methods which yielded different accuracies. Even with Object Oriented Classification suggested by Reda Yaagoubi you will need field data to "calibrate" or "validate" the allometrical models in order to have certain level of uncertainty. This is also true for several tree delineation approaches. If not field data you will need visual/manual delineation of tree crowns to train the detection algorithm whichever you decide to apply.

We worked with two oak species, and by the information you provide it seems that you may have a lot more species in urban environments. I suggest an alometric "model" per species or at least genus if possible.

Another issue in your excersise is the occurrence of trees in closed canopies such as urban forests or urban parks. This are quite a challenge.

Hope this helps, best regards and good luck

You may take a look to these:

Article Remote Sensing: An Automated Methodology for Olive Tree Dete...

Article Large Scale Palm Tree Detection in High Resolution Satellite...

Article Large-Scale Oil Palm Tree Detection from High-Resolution Sat...

But also

Thesis Automated tree detection: applications for object-based imag...

Conference Paper Individual Tree Detection from Multi-View Satellite Images

Conference Paper Voting based tree detection from satellite images

Conference Paper Automated tree detection and density calculation using unman...

Hope this helps

Thanks Alberto Gomez-Tagle, we will read those documents and think how to create or use a method that works for us.

My best regards.

Dear D. Object oriented segmentation with sub-meter resolution satellite imagery and using SW like eCognition sounds very promising! Good Luck!

Salem mohammad Ghaleb Issa Thanks for your recommendation. We are learning about these algorithm and how implement them.

My best regards.