Tree species classification using machine learning in Google earth engine

Summary

The conversation centers around a technical discussion on analyzing and classifying tree species using geographic data. One person shares a folder containing shape files and CSV files related to different tree species. They discuss the process of classification using Google Earth Engine and machine learning models in Python, focusing on optimizing accuracy. The conversation involves creating separate shape files for each tree species and preparing the data for classification. They agree to proceed with the task and exchange the updated files for further analysis.

Takeaways

• 😀 Discussion begins with greetings and confirmation of an email sent containing layers.
• 📂 The folder contains a boundary shapefile and a point shapefile with three species.
• 🗺️ The point shapefile contains data on tree species and is being used for classification.
• 🔍 The classification is to be performed using Sentinel imagery, which has a 10m resolution.
• 🌳 The user plans to classify all tree species together in one class for better accuracy.
• 🌲 The point shapefile contains six to seven different tree species, recorded in a 'tree type' column.
• 🏞️ The boundary and point shapefiles are not perfectly aligned in the visualization.
• 🤖 Machine learning methods in Python or Google Earth Engine will be explored for classification accuracy.
• 📊 The user is considering training samples using pixel values like NDVI and EVI for machine learning models.
• 📝 It is suggested to create separate shapefiles for each tree species to improve the classification process, which the user agrees to do.

Q & A

What type of files are being discussed in the transcript?

The transcript discusses a boundary shapefile and a point shapefile, with the point shapefile containing data on different tree species.

What software is being used to handle the shapefiles?

The speaker mentions using GIS software to open the shapefiles and later discusses using Google Earth Engine and Python for further processing.

What is the purpose of the point shapefile?

The point shapefile contains data on tree species and is used to classify different trees based on species for further analysis.

How many tree species are being classified in the shapefile?

There are six to seven tree species being classified in the point shapefile, with around 33,000 fields in total.

What image data is being used for classification?

Sentinel imagery is being used for classification, with the goal of achieving better accuracy. The resolution of the Sentinel image is mentioned as 10 meters.

What method is suggested for classification if Google Earth Engine doesn't provide accurate results?

If Google Earth Engine doesn't yield good accuracy, it is suggested to use Python and machine learning methods to classify the tree species.

What types of pixel values might be extracted for machine learning classification?

Pixel values such as NDVI, EVI, and other spectral indices might be extracted from the image data for machine learning classification.

What approach is recommended to improve classification accuracy?

It is recommended to create separate point shapefiles for each tree species and use them as training samples to improve classification accuracy.

What is the next step that the person will take to proceed with the task?

The person plans to separate the tree species into individual shapefiles and send them back via email to proceed with the classification task.

What is the overall goal of the process being discussed in the transcript?

The goal is to classify tree species using shapefiles, Sentinel imagery, and machine learning to ensure better accuracy in mapping the different tree species in the specified area.