Laser Sensing Display for UI interfaces in the real world
Fork of skinGames_forktest by
Scene.h
- Committer:
- mbedalvaro
- Date:
- 2013-10-30
- Revision:
- 44:2432c218f191
- Parent:
- 40:3ba2b0ea9f33
- Child:
- 45:5ef809480c12
File content as of revision 44:2432c218f191:
#ifndef SCENE_H #define SCENE_H #include <vector> #include "matrixClass.h" #include "classLaserSensingTrajectory.h" // BaseObjet use instances of the classes LaserPoint and LaserSensingTrajectory (cannot forward declare only) //#include "LaserRenderer.h" // Forward declared classes (attention: a pointer to an "incomplete class" - eg, only forward declared - is not allowed. It is just like unsing an actual object...) class LaserRenderer; class laserSensingDisplay; // A base class for an "object": // Note: all the 3d coordinates of the object are in PROJECTOR coordinates in fact... we don't store a POSE for each object, but instead the // final RT*vi where vi are the points of the objects in local coordinates. // Also, we store IN THE OBJECT the list of projected points (that is, the 3d points projected by using K). The object has methods to study this // list (because projected points are also "sensed") and get information about light or dark zones. // Now, to be able to test things (where the object was "touched", etc), we may need to mantain a correspondence between the projected points // and the original 3d point. There are two options: put them into a single structure "laserpoint", or use two separate arrays. class BaseObject { public: friend class Scene; friend class LaserRenderer; friend class laserSensingDisplay; BaseObject(unsigned int _id=0):myID(_id) {} ~BaseObject(); // Setters: void setID(int _id) {myID=_id;} void setColor( unsigned char _color) {myColor=_color;} // color is coded in three bits in fact (for RGB laser), or more or less... // Per-object adjustement of mirror parameters (could be automated if we know the number of points, etc. For the time being, this is manual): // Getters: int size() {return(vertexArray.size());} // ATTN: in principle, the vertexArray size and the myDisplaySensingBuffer number of points should be equal!! we could check this here? int color() {return(myColor);} int ID() {return (myID);} Box3d getEnclosingBox(); // Building methods: // NOTE: I will try to have an encapsulated class, so I prefer not to refer to global variables (like the renderer lsr) void addVertex(V3& _v3, Mat44& _RT); // this add a vertex USING THE CURRENT MODELVIEW (in global lsr object). NOTE: it is NOT projected and added to the myDisplaySensingBuffer yet. void clear(); // this deletes all the vertices and their projections from *this BaseObject instance void transform(Mat44& _RT); // this transform all the vertices of the object by _RT // void render(LaserRenderer* ptr_lsr); // this will project ALL the 3d points in the object vertexArray and store into display sensing buffer (myDisplaySensingBuffer), using // the current projection matrix and the current projecting mode in the global lsr object. NOTE: THE PREVIOUS DISPLAY BUFFER for this object is CLEARED. This ensures that when // "rendered", we have the same number of projected and 3d points // Sensing methods (note: only way to get data - touched, not touched, etc - through getters that will re-execute the processing on new data... // (we could have a boolean to check if new data is available, but since the displaying is done very fast, it is likely that we may need to re-compute // the processing anyway...) bool sense(void); unsigned char maxIntensity(void); unsigned char minIntensity(void); // Virtual methods? for more complex testing - like overall direction, or even motion in case of "sprites" // ... // NOTE: as always, I run into this annoying problem: if I have a vector of pointers to the base class, AND virtual methods, then I can store all the // childrens pointers in the vector (using downcast "dynamic type casting" which won't slice the object). But then, it means the base class declaration // needs to know somehting about the children methods! (at least their types and names). THIS IS REALLY ANNOYING, because I would like to add new kind // of children objects without needing to modify the base class declaration! How to do that? The thing is, sometimes I may need to call to a child method; // if the pointer points to an object that does not have this method implemented, NOTHING should happen (pure virtual function). Otherwise, it should work // as specified by this child. This very much reminds me to weakly typed Objective-C polymorphic arrays... (NSArrays). How to do this in C++? unsigned char myColor; // per object color (public because it will be used intensively by the displaying engine) LaserSensingTrajectory displaySensingBuffer; private: // PRIVATE INSTANCE VARIABLES: int myID; // Array of 3d coordinates and corresponding 2D projections with sense data: // These are private, so we can ensure that their size is the same... vector<V3> vertexArray; //LaserSensingTrajectory displaySensingBuffer; // the object displaySensingBuffer contains the trajectory of v2 projected points, as well // as some methods to perform the analysis of the saccade. NOTE: I could have made BaseObject a CHILD of the clas LaserSensingTrajectory. But I prefer to // have a clear separation between sensing parameters "tweaking" (like mirror delays, setting methods and variables) and the more pure graphical BaseObject Box3d enclosingBox; // this will be filled when rendering the object or when explicitly computing the enclosing box }; // ============================================================================================================================================== // The scene, which is just a collection of objets, plus some methods to add objects, etc: class Scene { public: friend class LaserRenderer; friend class laserSensingDisplay; Scene(): numTouchedObjects(0) {clear();} ~Scene(); // TO THINK ABOUT: METHODS TO TRANSFORM THE COORDINATES OF THE POINTS IN THE SCENE (affecting 3d coordinates) and to RERENDER, // or even TO TRANSFORM AND RENDER (but without modifying the original 3d points). // Create/destroy new objects: // TWO WAYS to create objects: // (1) Use a strategy somehow similar to OpenGL, by using a "begin/end" structure // (2) Making a specific object class inheriting from Object struct (or class), and then ADD to the Scene: void addObject(BaseObject* ptr_newObject); void addCurrentObject(); // this adds the "current" object (pointed by ptr_currentObject) void transform(Mat44& _RT); // this transform all the objects of the scene by _RT // Rendering whole scene: // void render(LaserRenderer* ptr_lsr); void clear(); // "delete" all the objects in the scene (attn! if we store the objects in the scene as POINTERS, we must 'delete' the pointer created with 'new', // so the memory for the object is liberated by calling its destructor, and THEN clear the vector of pointers) void deleteObject(int _id); // selective deletion (if we know the object index ; perhaps using a STL map with names?) int totalObjects(); // number of objects in the scene int totalPoints(); // total number of points in the scene int sense(); // Methods to "rearrange" the objects in the scene for rendering: // TO DO ........................ // Pointer to the current object being built (in case we use begin/end methods for creating and adding new objects to the scene). // In the future, we may have methods to set this pointer to some other object for adding vertices, changing color, etc. Perhaps not necessary... BaseObject* ptr_currentObject; // made it public because accessed by the WRAPPER FUNCTION "vertex" (can be made a friend...) //private: // The array of objects in this scene: vector <BaseObject*> objectArray; // better to use a vector of pointer: the advantage would be, we could have a polymorphic superclass "Objects" with virtual methods... //bool touchedScene; int numTouchedObjects; // only modified and accessed when calling sense() method }; #endif