I am working in QGIS 2.18 with thermal image raster data, and I would like to isolate from the total pixels in the image those that form a group of values higher than the average of their closest environment.

The basic idea is to identify those groups of pixels with high values, hotspots, to convert them to points.

In short, use a thermal image to identify hot spots automatically.

Any suggestions?

I did as @Kazuhito did but the result is not very encouraging. The idea is to find a process to identify hot spots in a mosaic of thermal images taken with an RPA.

Problem that I have in many cases, that the mosaic is burned on one side, very high values, or that very high values of temperatures are grouped, which can be rocks, bare ground, etc ...

If I run the local algorithm minimun and maximun of SAGA I get many records.

Would there be some way to preprocess the image to obtain better results?

  • 1
    What have you tried? See Extracting raster values and creating new raster in QGIS?: gis.stackexchange.com/questions/49800/…
    – BERA
    Oct 27, 2017 at 10:45
  • I tried to extract maximum values and later convert to points, but there are too many maximum value points, because one part of the image is hotter than another. To avoid this, the idea is to identify a grouping of pixels that surpass the average of their surroundings, corresponding to hot elements in the environment. Example animals in the field. I'm sure I should apply a filter, 4x4 or 8x8, but I'm not clear. Oct 27, 2017 at 10:52
  • Perhaps you could also have a look at other stackexchange sites. Image analysis and machine vision are common also in robotics, medical applications etc. and finding local maximums feels like rather basic task.
    – user30184
    Oct 27, 2017 at 13:29

1 Answer 1


You may be interested in a QGIS Processing algorithm SAGA | Vector <-> raster | Local minima and maxima tool. As you are interested only in maxima, please tick off minima.

(1) Left: random raster, Right: overlying maxima points

enter image description here

(2) Left: displayed only maxima points. Right: displayed the same points as heat map.

enter image description here

Probably this tool is rather new in processing... I used QGIS 2.18.13 with SAGA 2.3.2.

  • What is the criterion for calculation min and max local? Is tool using a default value for that (3x3, 4x4...)?
    – xunilk
    Oct 27, 2017 at 14:42
  • @xunilk So far I have not found documentation so honestly I do not know. My observation only; these distribution of maxima / minima points are not related to any window size (block or moving). Rather it seems conform to TPI. So I feel this tool is showing peak and pit derived from local convergence and divergence pattern.
    – Kazuhito
    Oct 27, 2017 at 14:57
  • OK. Thanks for your answer. It looks like max points are calculated based in cluster of pixels. So, if it is necessary to apply a 4x4 or 8x8 filter for determining these points, it can be used 'ReadRaster' GDAL method as block in a python code.
    – xunilk
    Oct 27, 2017 at 15:21
  • You can find the code here (but not sure it helps too much, definitely not a x by x grid though): github.com/johanvdw/SAGA-gis-git-mirror/blob/… Oct 4, 2018 at 15:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.