You could use the FeatureMerger (attributes only in parameters).  If there are small differences in time, round them off to the closest 15/30 seconds using the AttributeRounder.

---

I assume you have some ID?

 

 

But if you want that, just merge them (suplier = 1 and requestor = 1) all and calculate difference.

 

Then listbuilder grouped by ID, listsorter on difference ascending and listindexer with index=0.

---

Hi,

 

 

Another thought.

 

If you replace the images and GPS points with time positions on a time line, the NeighborFinder can be used to find GPS point which has minimal time difference for each image.

 

Assuming that the format of datetime is "yyyymmddHHMMSS".

 

 

(1) Use a RasterExtractor to store each raster (image) data as an attribute.

 

(2) Replace each raster with a point on the time line by a VertexCreator.

 

X: datetime of the image, Y: 0

 

(3) Extract the coordinate (x, y) of each GPS point by a CoordinateExtractor.

 

(4) Replace each GPS point with a point on the time line by a VertexCreator.

 

X: datetime of the GPS point, Y; 0

 

(5) Add a NeighborFinder; send the (2) points to the Base port; send the (4) points to the Candidate port. Matched features will have the coordinate (x, y) of the closest candidate (GPS) point on the time line.

 

(6) Finally use a RasterReplacer to restore the original raster.

 

 

Alternatively, (assuming the rasters have unique ID attribute) you can use a FeatureMerger to merge the Matched features to the original images, instead of the RasterExtractor/RasterReplacer pair. 

 

 

Takashi

---

If you adopt the time line method, it's better to format the datetime with "%s" (epoch time), rather than "yyyymmddHHMMDD".

---

Time line method works perfectly, thank you Takashi

 

At first sight, both date/time formats work well for this method

---