I think you must combine differents parameters such NDMI, NDRE, MSAVI, RECI

Then look for an algoritm

NDVI is a reflectance-based index used to analyze vegetation vigor. It may be able to differentiate crops based on their unique spectral signatures. However, an important consideration is whether the spectral and spatial resolutions of the remotely sensed data, from which NDVI is derived, are adequate for this purpose. I recommend starting with an investigation at known locations to determine if NDVI values vary between different crops at those sites. If variations are observed, you could proceed with a supervised classification on the NDVI image by using 'training' signatures from these known sites. Compare the outcomes of the supervised classification and refine the process as needed by incorporating additional sets of signatures for various crops from known locations.