If the string is always longer than 2 numbers, then the regex is simple. Using GREP, this works:
(Assumes the desired string is always at least 5 numbers---"num" is the file I generated to hold your example)
Code:
grep -o '[0-9]\{5,\}' num
The regex translates to "5 or more instances of any digit". The "o" flag tells grep to return only the matched pattern.
This can also be done with SED, but probably will be messier