Postclassification of OBIA classification using GRASS and PostGIS

This repository contains a jupyter notebook which allow to perform postclassification of a land cover map using GRASS GIS and Postgis.

Cite this code

Please use the following DOI for citing this code:

How it works ?

The notebook enables for computation of neighborhood matrix on the segmentation result, using r.neighborhoodmatrix. Then, several table manipulation are made using PostGis in order to obtain a final table with the following informations for each segment:

• The label of the segment
• The proportion of border shared with the different classes
• The label of the classes sharing the first, second and third more important portion of the border
• Shape statistics of the segment
• Aggregated statistics values from another raster (e.g. spectral value, NDVI, nDSM)

Example of the table

seg	label	prop_11	prop_13	prop_14	prop_20	prop_30	prop_31	prop_41	prop_51	first_label	second_label	third_label	area	perimeter	compact_circle	compact_square	fd	ndsm_min	ndsm_max	ndsm_mean	ndsm_stddev	ndsm_median	ndvi_min	ndvi_max	ndvi_mean	ndvi_stddev	ndvi_median
2767795	11	0.5959	0.0685	0.0000	0.3219	0.0000	0.0000	0.0000	0.0137	11	20	13	192	146	2.9723	0.3796	1.8958	-0.1790	2.0029	0.8118	0.7412	0.8244	-0.0152	0.3316	0.1104	0.0700	0.0963
3076490	11	0.3929	0.1964	0.0000	0.3750	0.0000	0.0357	0.0000	0.0000	11	20	13	65	56	1.9594	0.5759	1.9286	0.1255	2.0264	1.0716	0.5021	1.1273	0.0280	0.2473	0.1260	0.0491	0.1297
3297859	11	0.4462	0.2000	0.0000	0.3538	0.0000	0.0000	0.0000	0.0000	11	20	13	250	130	2.3194	0.4865	1.7631	-0.0016	2.7471	1.1398	0.8502	1.1866	-0.0352	0.1996	0.0734	0.0413	0.0801
2120483	11	0.0455	0.0000	0.0000	0.4545	0.5000	0.0000	0.0000	0.0000	30	20	11	16	22	1.5515	0.7273	2.2297	0.6039	1.3888	1.1740	0.1747	1.2379	0.0473	0.2025	0.1234	0.0493	0.1224
2120882	11	0.2391	0.0000	0.0000	0.3261	0.4348	0.0000	0.0000	0.0000	30	20	11	88	46	1.3833	0.8157	1.7102	-0.0305	0.3824	0.0150	0.0678	0.0000	-0.0135	0.3317	0.1105	0.0744	0.1011

Example of reclassification

Optical image

Initial classification (OBIA + Random Forest)

Final post-classification (public release map)

Related code (OBIA)

If you are interested in Object-based image analysis and classification, take a look at our work:

• Opensource_OBIA_processing_chain
• Semi_automated_OBIA_processing_with_local_USPO