Intelligent Bill Scanning Solution

A long retail receipt does not fit inside one camera frame. Capturing it in pieces is easy; turning those pieces into one readable image without freezing a Flutter interface is the real problem.

Moving the expensive work across the right boundary

The prototype captures overlapping frames and passes image preparation, alignment and stitching to OpenCV 4.10 in C++. Dart FFI provides the bridge, while a background isolate prevents native processing from blocking camera feedback or the rest of the interface.

The native code is packaged as a Flutter plugin rather than hidden inside one application target. Setup scripts prepare headers, frameworks and platform libraries for Android and iOS, with documented paths for Windows and macOS. That made it possible to change the stitching algorithm without rewriting the Dart interface.

What the prototype proved

The application combines camera and gallery input, sensor and permission handling, BLoC state and a zoomable result view. More importantly, it established a reusable pattern for work that Flutter should coordinate but should not execute on its main isolate.

The useful outcome was not merely one stitched receipt. It was evidence that a mobile product could keep a responsive Dart experience while delegating a heavy computer-vision task to native code.