A slight variation on toolic's and eyepopslikeamosquito's solutions in that I use a closure to generate the odd numbers (it returns the next odd number every time it is called) and I wrap the logic for finding duplicates in a do block to avoid leaving the %seen hash lying around after it is no longer needed.

$ perl -Mstrict -Mwarnings -MData::Dumper -e '
> my $oddNos = do {
>    my $val = -1;
>    sub { $val += 2; };
>    };
> 
> my @array = qw{
>    M94202
>    M94150
>    M94297
>    M94150
>    M94161
>    M94161
>    M94162
>    };
> 
> @array = do {
>    my %seen;
>    map {
>       exists $seen{ $_ }
>       ? $seen{ $_ }
>       : do {
>            $seen{ $_ } = $oddNos->();
>            $seen{ $_ };
>            };
>       } @array;
>    };
> 
> print Data::Dumper->Dumpxs( [ \ @array ], [ qw{ *array } ] );'
@array = (
           1,
           3,
           5,
           3,
           7,
           7,
           9
         );
$
[download]

I hope this is of interest.