Converting from raster to vector in FME is notoriously slow.
You are on the right track for ways to improve performance. If you use the RasterExpressionEvaluator to create a 1 band image where your yellow RGB value(s) have a value of 1, and everything else 0, and then set the Nodata Value to 0 (RasterBandNoDataSetter) then the coersion should be more efficient.
You can also experiment to see of the RasterExtentsCoercer (DataExtents) or RasterToPolygonCoercer is more efficient than the RasterCellCoercer for your particular data.
Hi @danilo_inovacao Yep you will see a vast improvement in perforamnce if you do as @jdh suggests. For your particular case you can use this expression in the RasterExpressionEvaluator (use Int8 as the
interpretation):
@if(A[0]==255 && A[1]==251 && A[2]==134,1,0)
Follow this with a RasterBandNoDataSetter to set 0 = NoData and then a RasterToPolygonCoercer to get it into vectors and Bob's your uncle! The process shouldn't take any longer than 5 seconds!
The output should be a line or a polygon?
Anyway, I have experienced that a 'Generalizer' after this raster-to-vector conversion can be helpful in further processing. Just saying!
Converting from raster to vector in FME is notoriously slow.
You are on the right track for ways to improve performance. If you use the RasterExpressionEvaluator to create a 1 band image where your yellow RGB value(s) have a value of 1, and everything else 0, and then set the Nodata Value to 0 (RasterBandNoDataSetter) then the coersion should be more efficient.
You can also experiment to see of the RasterExtentsCoercer (DataExtents) or RasterToPolygonCoercer is more efficient than the RasterCellCoercer for your particular data.
Thanks @jdh . Your suggestions was very important for me
The output should be a line or a polygon?
Anyway, I have experienced that a 'Generalizer' after this raster-to-vector conversion can be helpful in further processing. Just saying!
Hi @jneujen, the output is polygon.
Thanks your help :)