hello,
How can determine the number of sampling points in field for check the classification of satellite images?
what is your question is ?
what you want to know about?
Explain
if the meaning of number of sampling points is the types of classes you want to make in your classifications of satellite images, it may depend upon what is your land cover, what kind of problem you are dealing with in hand? if this classification of vegitation, then you have to keep in view slope, aspect, season..etc/
Tell me a bout what kind of classification you are doing?
I used inter class variability in my urban change detection. Paper is here
Urban change detection of Lahore (Pakistan) using the Thematic Mapper Images of Landsat since 1992-2010
You may want to read this.
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwjY9d2tqs7NAhUJ8mMKHYjUCzAQFggkMAE&url=http%3A%2F%2Fwww.mdpi.com%2F2072-4292%2F6%2F2%2F964%2Fpdf&usg=AFQjCNE1sEEjwLwxKTKdKQGxO8MYi9pVIQ&sig2=PDMBRRirtE8Bjgm8VE1-AA
Whether you want to use those points for the classification itself or to check for accuracy of the classification you can use this suggestion;
1a. A feature can be very large in an area and all u will need is just a sample especially if its homogeneous. This means you may need a sample each for each feature.
1b. for heterogeneous features, this could be challenging but you could make a grid of the study area, and locate points on the image that are confusing and go to the field with those point to locate the type of feature in those places.
2. Remember the objective of classification is to make sure you are as close as possible to what's on the ground and getting field data though is very important it shouldn't replace REMOTE SENSING which is the fundamental objective of a satellite image.
I hope this helps.
How namber do sample need? so 10 % pixles or point in field
are there methods for sample size
Please go through the following research paper. I am sure that all your queries has been answered in this publication. You can download the full research paper from my researchgate website at https://www.researchgate.net/profile/Hari_Srivastava2
Patel, P. and Srivastava, H.S. (2013), Ground truth planning for synthetic aperture radar (SAR): Addressing various challenges using statistical approach, International Journal of Advancement in Remote Sensing, GIS and Geography, Vol. 01, Issue 02, pp. p. 1-17
Hi!
In an qualitative approach, you can take a sampling point in the center of each temathic class under different conditions (in this way, the number of sample points will be directly proportional to the number of classes you need to verify).
In an quali-quantitative approach, you can use type-sections (like a topographic profile) to take sampling points in every significant change you identifiy (you need to choose representative type-sections of the phenomenon that you're studying).
Also, you can use a grid (equal areas) to collect sample points in each portion. The grid size will be related to the map scale and the phenomenon you are studying.
Sometimes a pure quantitive approach are not suitable, because you can miss something important or collect to much repeated data. However, i use 10% of the sampled points for validation.
I hope this can help you.
Best regards...
Dear Bibi Zahra,
More training samples are usually beneficial, as they tend to be more representative to the class population, a small number of training samples is obviously attractive for logistic reasons.
It is often recommended that a training sample size of maximum likelihood classifier for each class should not be fewer than 10–30 times the number of bands.
For more details, please go through the following research paper:
Comparison of Classification Algorithms and Training Sample Sizes in Urban Land Classification with Landsat Thematic Mapper Imagery.
Best Regards,
Saati
Article Comparison of Classification Algorithms and Training Sample ...
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy