Regenerative agriculture certification requires rigorous and periodic inspection of adoption of recommended principles on the field. One of the key principles of Regenerative agriculture is to follow multi-cropping, which results in conservation of biodiversity. In our study, we demonstrate a method for detection of multi-cropping in a farm using remote sensing data and machine learning based models. We define classification problem with 4 classes: A. Mono-cropping - single crop fields, B. Mono-cropping - single tree-species plantations, C. Multi-cropping - single crop with trees, and D. Multi-cropping - agroforestry with multiple crop and multiple tree species. For all these classes, ground truth data has been created through manual annotation of the fields. Remote sensing satellite images of 10m resolution from Sentinel-2A have been used as an input data for classification. We present the results obtained by experimenting with Support Vector Machine (SVM) and Random Forest (RF) models with manual and AutoML based training. Using the proposed approach, the number of visits for assessment of regenerative agriculture compliance can be reduced significantly, which will further reduce the certification cost and hence the overall cost of agricultural produce.