Using cameras

Cameras are used to capture your data. There must be at least one camera in your scenario in order to get data out but any number of cameras can be placed. Each camera will produce one complete set of images, meta data and annotations that will be output to a folder in the output hierarchy. This folder will be named after the database path name you give to the camera instance when it is placed in the scenario. The data in the output folder is rendered from the perspective of its camera and the world space location of the camera is stored in the annotations file so that cross referenced multi-camera setups such as stereo vision can be simulated.

Cameras are assets that are imported to chameleon in .mtb files just like standard assets and appear in the cameras list in the editor. Camera assets can be set up to simulate real life camera properties such as color and noise profiles. (Most camera properties such as field of view and capture format are set up in the editor).

Cameras can be placed directly into the scenario by selecting a camera asset from the menu and placing it, or they can be attached to other elements in the scenario by selecting it via that element. Cameras that are attached to a moving object will move along with it, keeping the fixed relationship established when it was attached.

Cameras are always loaded as a standard type and can be set to have certain behaviors and attributes by changing the selection in its camera type dropdown. Below describes each of Mindtech's camera types and their uses.

The standard camera produces a rectilinear image using a simple pinhole model. Most of the functions of this camera are shared with the rest of the camera types.

Setting camera type to drone camera causes the camera to behave as though it is attached to a drone. To use a drone camera you first need to create a circuit, you can find how to do that in Using Circuits and attach the drone to it by passing the name of the circuit in the 'Circuit' field (Fig.2 - 1).

When the simulation runs, the camera will fly along a straight line between the waypoints of the circuit, at a fixed height above the waypoints according to the Height setting (Fig.2 - 2) and at a constant speed according to the Speed setting. Note that the height of each waypoint can be set individually when creating the circuit, allowing for variations in drone height above the ground.

The camera position may be set up to lie on and align with the circuit at the start of the simulation but it does not need to be. On startup, it will automatically find the nearest waypoint, set its height to the waypoint height plus the value set in the Height parameter and fly towards it. The facing direction will be automatically set to point to the waypoint unless it has been set to not do that as described later.

When the camera arrives within a certain radius of the waypoint, it will select the next point in the circuit and fly towards it. Note that if the circuit is composed of waypoints arranged as a net work rather than as a single loop, there may be waypoints that have multiple possible next connections in which case the next waypoint will be chosen randomly from the list of candidates. The choice is controlled by the Seed value.

The drone camera has three look at modes, when Fix Direction (Fig.2 - 4) is set it will always point in the direction set in the scenario editor otherwise it will always point towards the next waypoint with a smooth transition when the direction of flight changes. If the drone camera is given a target object, it will always point towards the center of that object.

Setting camera type to cloud camera causes the camera to relocate after a set number of images have been captured and continue capturing from the new location.

Unlike a standard camera, the cloud camera is not located at the placement point in the scenario, (the location defined by the camera position and rotation fields). In cloud camera mode this point defines the center of a hollow sphere, or cloud, within which the camera will appear. It is possible to set the inner and outer radius of the hollow sphere and the camera can be restricted to a smaller volume by setting the various parameters.

It is easiest to think of camera placement in terms of longitude and latitude. The spread restricts the camera to locate itself within a certain range of longitude while the direction defines the center of the excluded area. The max and min elevations set the limits of the latitude where the camera may locate. All locations are relative to the center point; note that the camera is aware of other objects in the scene and will not locate itself inside any object although it will place itself in locations where the target may be occluded, including below ground. so caution is needed in setting possible placements.

The camera will be aimed at a random point within a target volume which is by default 1 meter on a side and located at the center of the sphere. The location and size of the target can be changed without affecting the location of the sphere. To ensure the camera always points at the same spot, set the dimensions of the target to zero.

During simulation runs the annotation file for a cloud camera records the current location and orientation of the camera, not the location and orientation of the cloud sphere.