You may not be interested in a little reading on seismic convolution. This is a process to recognize lines (horizons, changes in the propagation of sound waves causing a reflection) in seismic data. Seismic recordings of reflections (echoes) from a source (sound generator) to geophones (to record the echoes) show up as hyperbolas if plotted. Unfortunately the data from multiple sources do not line up. To combine data from multiple sources the reflection times are migrated to make everything line up. This sounds like what you are trying to do as you want to recognize lines. This technique goes a little farther as it handles lines of any angle.
The calculations involved take a lot of computing power and are done on supercomputers in the oil industry, but are well within the reach of today's personal computers. On Wed, Feb 19, 2014 at 5:25 AM, Henry Rich <[email protected]> wrote: > Look at > > (1 10$0) E. image > > or > > (10 $ 0) E. line > > Beware of a previously-reported bug in non-literal E. > > Henry Rich > > > On 2/19/2014 7:04 AM, Joe Bogner wrote: > >> I can get it aligned, but I'd like to flag the start of the line and not >> just when I first detect a line. >> >> For example, I'm saying a line is when 10 consecutive pixels are 0. This >> flags the 10th pixel. When I reshape and align, I'm losing the line >> starts. >> I apologize for the poor wording of this. Perhaps a picture and small test >> case will help: >> >> http://imgur.com/tvQn6ZD >> >> ]lines=: 2 22 $ (0 0), (10 $ 1), (10 $ 0),(5 $ 1), (10 $ 0), (4 $ 1), (6 $ >> 0) >> ]hlines=:(5 (+/)\"1 lines) = 5 >> ]hlines2=:2 22 {. hlines >> >> >> (note: the 2nd row only has 1 line since the 2nd block only has 4 >> consecutive pixels = 1) >> >> lines >> 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 >> 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 >> hlines >> 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 >> 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 >> hlines2 >> 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 >> 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 >> >> I would like hlines2 to be: >> >> hlines2 >> 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 >> 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 >> >> So in this case, I find the 1s and make the next 4 pixels 1. I'm not sure >> how to go about that or if this is the right way to solve the problem. >> >> I might be able to trigger the start with Outfixes... <\."1 lines, but >> that >> seems inefficient and only gets me partly there. I only considered a state >> machine but that also seems heavy. Thanks for any input >> >> >> >> On Tue, Feb 18, 2014 at 9:42 PM, Joe Bogner <[email protected]> wrote: >> >> Raul, thanks! Yes, that solved the mismatched shape problem. Now I just >>> need to tinker with getting it aligned. >>> >>> On Tue, Feb 18, 2014 at 9:20 PM, Raul Miller <[email protected] >>> >wrote: >>> >>> It looks like you are losing 9 pixels from each array. >>>> >>>> The easy answer would be to put them back. Ideally, you would try to put >>>> some back on each side so that the result is not too far off center. >>>> >>>> Thus: >>>> hlines2=: 58 157 {. 58 _152 {. hlines >>>> vlines2=: 58 157 {. _43 157 {. vlines >>>> >>>> Note that your matched content will be smaller than the original. That's >>>> also fixable if it's an issue for you. >>>> >>>> >>>> >>> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm >> >> ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
