Solving the Issue of multi-band aerial images for the application of deep learning models

Mona Younis

????? ??????? ?? ????? ????? ?????? ???????? ???????? ?????? ArcGIS Living Atlas of the World ????? ?????? ?????? ??????? ??

?? ?????? ?? ???????? ???? ????? ??? ????? ?????? ?????? ?????? ????? ? ????? ?????? ??????: ??????? ??????? ???? ??? ???? ?????? ???

(RGB)? ?? ??? ?? ????? ?????? ???? ?? ?????? ???? ????? ?? ?????? ?????? ?? ????

( ??? NIR - ?????? ??? ???????).

In many projects that rely on deep learning (deep learning) models for image analysis, we face a common challenge: requests for only three-band images (RGB), while the aerial images that are often available are four or more bands (such as NIR - Infrared).

This affects the accuracy of the model when applying the model directly. I decided to share the solution that succeeded in this problem:

???? ???? ??? ????? ??? ??????? ??? ????? ??????? ???? ????? ????? ???? ?????? ???? ???? ??? ?? ??? ???????

Solving ..

??? ???????? ????????: ??????? ????? ????? ???????? ????????: ??????

(Red)? ?????? (Green)? ??????? (Blue).

Raster in ArcGIS Pro ??????? ???????? ???????? : ???????? ?????

From the raster functions, choose extract Bands.

Enter input Raster and in Combination number Bands and in Missing Band choose Fail, then press Create new Layer

Export Raster from Output Extract Functions

choose Output Location and Run

From Layer Properties in Source, the number of Bands became only 3.

?? In a world increasingly reliant on artificial intelligence and geospatial data analysis, understanding how to address technical challenges like these is not just a skill—it's a gateway to innovation. ??