Most probably your scan contains only images, not text. However, without machine-readable text ResearchGate's algorithm for identifying references cannot work, because different to a human it is not able to read text in an image. There are programs that allow to recognize the text in the images ("OCR") and to store it in the background of the PDF file. However, my experience is that ResearchGate's algorithm works only once, i.e. even when you upload a new version of the the file it will probably not check for references again. - In similar cases in the future, you may try to use the method developed by Andrejs Krauklis : https://www.researchgate.net/post/Method_for_Improving_Recognition_of_Citations_References_on_ResearchGate/1. Create a separate file with numbered references and upload it as a private file, prior to uploading the main file as a public one.