Unfortunately, the available information used to generate NCHILD in MTUS does not allow for a distinction of own children. Additionally, the child identifier variables in AHTUS may not necessarily only identify own children, and also could identify any child living within the household. With that said, as you point out, strictly identifying own children, for years 2003-2018, is a nice feature of the ATUS data (see here).
Similar to your previous question, it seems there are two possible ways to proceed. The first approach is to merge the MTUS data onto the ATUS and AHTUS data. You can match on CASEID in ATUS and PERSID in MTUS, when the data are restricted to respondents only. You can match on IDENT and SAMPLE, which are available in both MTUS and AHTUS. The second approach is to just use the information in the ATUS and AHTUS files, if you do not need any additional information from the MTUS files.