For my own needs, I wrote a little program to read the zip files, reorganize the data in a tree structure, and generate a json file or directly populate a database (a specific class must be written to it).
The application can be launched in command line, or be integrated in a larger application.
It is published under MIT license, here: https://github.com/inrae/phpodkread.
So this is a bit like a command-line ODK Briefcase that allows to take the survey data directly from ODK Collect without a server in between?
Personally, I don't know the format generated directly by ODK Collect: I have access to an ODK Central server, and I retrieve the generated zip files. What the script does is to read the zip files that contain the data, restructure them, and the result can be used to populate a database by simply creating an appropriate class.
The script can be integrated into a PHP application quite simply, the "engine" being encapsulated in separate classes and files (lib folder).