As someone who has created many puzzles based on combining two different puzzle types, here is one key piece of advice that you may not be hoping to hear...
Not all puzzle types can be paired up well in this way.
Here are a few issues with attempting to combine slitherlink and nonogram:
As you already point out, many nonograms require numbers greater than 3. It is important to note therefore that any nonogram requiring four or more consecutive shaded cells in a row will not equate directly to slitherlink clues.
With slitherlink puzzles, the numbers are already provided at the beginning. You are not trying to generate numbers, rather you are trying to create a line. If you wish to use the slitherlink keys for your nonogram row indicators, the solver would not even need to solve the slitherlink, making it a redundant puzzle feature - what you would have is just a set of numbers that could separately be both a slitherlink and the row indicators for a nonogram. This would just create a puzzle whose solution path spreads out in different directions rather than converging on a single answer.
Both of these puzzle types suffer from problems with ambiguities, making it very difficult to create such a puzzle while ensuring a unique solution exists for both. As a trivial example, in your example slitherlink above the entire bottom row of squares is redundant (the line does not need to dip two units below the keys provided). Infamously, nonograms requiring two isolated diagonal shaded squares cannot be uniquely defined:
This point in itself does not mean it is impossible to create a unique solution for both, just that you will have to take extra care (and probably a lot of extra time) to ensure both solutions are unique, otherwise you will have a bad grid-deduction puzzle on your hands.
The key to producing a combination grid-deduction puzzle where the final step is a nonogram is to work backwards from the intended nonogram solution. After all, the reward of solving a nonogram is to find some kind of meaningful recognisable image, like a picture, a set of words, or a visual cipher (e.g. pigpen) - it just isn't enjoyable to solve a nonogram where the solution looks like static on a screen. You will therefore need to create your desired nonogram first, taking care to ensure no numbers greater than 3 are required in the row indicators. However, there is no guarantee that your target nonogram solution will generate numbers that also serve as keys to a uniquely solvable slitherlink.
Even if you do manage to create a compatible nonogram, the resulting slitherlink may be heavily over-clued, making it almost trivial to solve.
As these points hopefully make clear, combining these two puzzle types well is not going to be easy. Also, as the issues in Point 2 suggest, we're really going to need to find a way to produce numbers from the slitherlink that are not just its starting keys that are already provided from the off...
All of which leads me to a second (more reassuring) piece of advice:
Expand your thinking to other ways you can encode information in a specific puzzle type.
Just because a puzzle type relies on certain starting numbers doesn't mean those are the only numbers you can use for your target encoding.
Here's one idea: Use certain unclued squares in the slitherlink to be your nonogram row indicators. Shade these squares a different colour; instruct that the number of line segments around these squares yields the target number. The user must then actually solve the slitherlink to generate these.
Trivial example: Let's say the target nonogram is this:
A suitable slitherlink for generating the row indicators would then be something like:
Note that this approach could even lend itself to generating numbers greater than 3 if you use colour-coding. Using a natural ordering like colours of the rainbow, you could say that in each row you need to find 'the total number of line segments surrounding individual squares of the same colour'. Ordering them in rainbow order (ROYGBIV) then gives you the specific ordered arrangement required for the nonogram.
In conclusion, this will not be easy - but it may just about be doable and produce a meaningful combination of the two puzzle types if you use unclued slitherlink spaces for your targets.
Remember that not all nonogram solutions will lend themselves to an easy puzzle creation process. It is also important to note that the larger your nonogram, the larger your slitherlink will be! It might be best to aim to create a fairly small puzzle in this way before attempting anything larger, or you might find yourself stuck while creating it! Good luck :)