Something like this perhaps?

#!/usr/bin/env perl 
use strict;
use warnings;
use Test::More;

my @rows = (
        {
                have => 'V12345:name_test',
                want => 'V12345:name_test'
        },
        {
                have => 'V12345_name_test',
                want => 'V12345:name_test'
        },
);

plan tests => scalar @rows;

for my $row (@rows) {
        my $out = $row->{have};
        $out =~ s/(\d)_/$1:/;
        is ($out, $row->{want});
}
[download]

How well this regex works depends on how true your sample dataset is. You can amend @rows and the regex until you have covered all the possibilities you require. See also How to ask better questions using Test::More and sample data.

Addendum: I see that toolic has used precisely the same substitution in his answer while I was constructing this. That bodes well for it as a solution to your specific question.