Skip to content

Latest commit

 

History

History
27 lines (20 loc) · 2.21 KB

README.md

File metadata and controls

27 lines (20 loc) · 2.21 KB
Raw

ARTrade

A simple Android app created for the January 2018 OANDA hackathon. The app utilizes the ARCore SDK and detects flat surfaces as planes. It then lets you place a 3D object on a detected plane and interact with it.

Inspiration

The original idea was to develop a mobile app that lets you interact with a static AR target and make a Forex trade. This app is an MVP which demonstrates how the original idea could be fully realized with some modifications. Currently it has the essentials to achieve this:

  1. A way to place a 3D object on a surface detected from a camera feed
  2. Logic to detect that the object has been touched (the ARCore Java SDK does not include this functionality out of the box at the time of writing)

Implementing Object Detection

The following code snippet inside ARActivity.java is used to detect when an object has been tapped on the screen:

  Pose objectPose = mAnchors.get(0).getPose();
  Pose currentPose = hit.getHitPose();
  double distance = Math.sqrt(
          (currentPose.tx() - objectPose.tx()) * (currentPose.tx() - objectPose.tx()) +
                  (currentPose.ty() - objectPose.ty()) * (currentPose.ty() - objectPose.ty()) +
                      (currentPose.tz() - objectPose.tz()) * (currentPose.tz() - objectPose.tz())
  );

  if (distance < 0.1) {
      ...
  }

The above approach aims to calculate the distance between a point in space realized by a tap (stored in currentPose) and a point anchoring the 3D object (stored in objectPose) and decide if it is small enough to be considered within the boundaries of the actual object (the condition distance < 0.1, where units are in meters). This is a very crude approach, since it models the touchable area of the object as a sphere with a radius of 10 cm around the anchor point. Due to the limited object detection capabilities of the ARCore Android SDK, this is about the simplest way of realizing it without introducing external graphics libraries or writing a full-fledged raycaster.

N.B. The Pose object is modeled as a quaternion rotation followed by a translation vector, so to get the translation components, we use tx(), ty() and tz() instead of qx(), qy() and qz().