I'm assembling a new genome for a plant de novo. I already have the scaffolds and most of the assembly but the parts in the centromere regions won't align against the reference sequence i have of the plant. This reference sequence is not of good quality. Is there a tool to get a good assembly of the centromere regions or an other way to get the centromere regions right?

This is hard. Use very long reads eg Nanopore, and also Pacbio. Use a very high coverage (the human T2T project used 170X of both if I remember rightly). Employ centromere experts and a large multinational team with extensive funding.

You can see this isn't trivial right ?

Why don't you start by using a telomere finder like tidk https://github.com/tolkit/telomeric-identifier to discover if the telomeres are all present ?

Also, you'll need to give details. Is this a small plant <1 GB, or a massive polyploid, or something inbetween ?

I haven't found a good tool to detect centromeres yet, but would be interested in your approach.