A canonical URL is the address a search engine picks as the representative version of a page that exists at several URLs. Think of a product page that is also reachable with a sort parameter, or over both http and https. Google groups such duplicate content and picks one URL to show. You can state your preference; Google calls that a hint, not a rule.
Example
<link rel="canonical" href="https://www.example.com/bikes/city-bike-black">This line goes in the <head> of every variant of the page, including the preferred version itself. According to Google, a permanent redirect and a rel="canonical" are strong signals; listing a URL in the sitemap is a weak one.
Common mistake
A canonical that points to a page that does not work: an error page, a redirect or a page with noindex. Relative addresses, and different canonicals for the same page through different methods, also blur the signal. Google advises against using robots.txt or noindex to choose a preferred version.
How do you check it?
Search the page source for canonical and open the address it gives: does it return a 200, and may it be indexed? Atlas, the research engine developed by Outseen, checks canonicals the same way and looks at what the target page actually returns. More technical checks are in the technical SEO checklist.