Source code for

import logging; logger = logging.getLogger("morse." + __name__)
from morse.core import blenderapi
import morse.core.sensor
from morse.helpers.components import add_property

[docs]def copy_pose(obj_from, obj_to): obj_to.worldPosition = obj_from.worldPosition obj_to.worldOrientation = obj_from.worldOrientation
[docs]class Camera(morse.core.sensor.Sensor): """ A generic camera class, which is expected to be used as a base class for real camera. Concrete instantiation are currently: - :doc:`video_camera <../sensors/video_camera>` - :doc:`depth_camera <../sensors/depth_camera>` - :doc:`semantic_camera <../sensors/semantic_camera>` .. note:: The cameras make use of Blender's **bge.texture** module, which requires a graphic card capable of GLSL shading. Also, the 3D view window in Blender must be set to draw **Textured** objects. .. note:: The streaming of data from this sensor can be toggled off and on by pressing the SPACE key during the simulation. This will affect all the video cameras on the scene. Toggling off the cameras can help make the simulation run faster, specially when there are several cameras. However, the lack of data on the stream may cause problems to some middlewares. .. note:: The **cam_focal** and **cam_fov** properties are linked together. Blender automatically computes one when setting the other. Therefore when setting **cam_fov** to something other than None, the **cam_focal** parameter is ignored. .. warning:: Contrary to most of objects in Morse, the X axis of the camera is not "in front" of the camera. Here, Morse follows the "standard convention for camera", i.e. X and Y are in the image plane, and Z is in the depth axis of the camera. """ _name = "Generic Camera" _short_desc = "Base class for cameras in MORSE" # Set the values of image size from the variables # in the Blender Logic Properties add_property('image_width', 256, 'cam_width') add_property('image_height', 256, 'cam_height') add_property('image_focal', 25.0, 'cam_focal') add_property('image_fov', None, 'cam_fov') add_property('near_clipping', 0.1, 'cam_near') add_property('far_clipping', 100.0, 'cam_far') add_property('vertical_flip', True, 'Vertical_Flip') add_property('retrieve_depth', False, 'retrieve_depth') add_property('retrieve_zbuffer', False, 'retrieve_zbuffer') def __init__(self, obj, parent=None): """ Constructor method. Receives the reference to the Blender object. The second parameter should be the name of the object's parent. """"%s initialization" % # Call the constructor of the parent class morse.core.sensor.Sensor.__init__(self, obj, parent) # Set the background color of the scene self.bg_color = [143, 143, 143, 255] self._camera_image = None """ Check if the bge.render.offScreenCreate method exists. If it exists, Morse will use it (to use FBO). Otherwise, Morse will se the old and classic viewport rendering. It requieres creation of additional scenes for each camera resolution, and synchronisation of these scenes before rendering (which makes the process potentially slower and may introduce some glitches if some objects are not properly synchronised). Unfortunabely, it seems that the current FBO handling do not work properly for Z-Buffer. Turn off temporarly until we find a solution (on Morse side), or the issue has been fixed (if needed) on Blender side. """ #self._offscreen_create = getattr(blenderapi.render(), 'offScreenCreate', None) self._offscreen_create = False if not self._offscreen_create: self.scene_name = 'S.%dx%d' % (self.image_width, self.image_height) persistantstorage = morse.core.blenderapi.persistantstorage() parent_name = is_parent_external = False for robot in persistantstorage.externalRobotDict.keys(): if == parent_name: is_parent_external = True break if not is_parent_external:"Adding scene %s" % self.scene_name) blenderapi.add_scene(self.scene_name, overlay=0)'Component initialized, runs at %.2f Hz', self.frequency)
[docs] def default_action(self): """ Update the texture image. """ # Configure the texture settings the first time the sensor is called if not self._camera_image: if blenderapi.isfastmode(): logger.warning("Running in fastmode! No camera support!") else: # Prepare the camera object in Blender self._setup_video_texture() # Exit if the cameras could not be prepared if not self._camera_image: logger.warning("Blender's bge.logic does not have the 'cameras' variable, \ something must have failed when configuring the cameras") if self._camera_image: if not self._offscreen_create: # Update all objects pose/orientation before to refresh the image self._update_scene() # Call the bge.texture method to refresh the image self._camera_image.refresh(True)
@property def image_data(self): if self._camera_image is not None: return self._camera_image.source logger.debug("image_data not yet available") return None def _update_scene(self): for _to, _from in self._scene_syncable_objects: try: copy_pose(_from, _to) except Exception as e: logger.warning(str(e)) def _compute_syncable_objects(self): """ Compute the relation between objects in the current scene and objects in the main logic scene. The logic is a bit complex, as in the case of group, we can have objects with the same name (but different ids). So, in this case, we follow the hierarchy on both scene to find correspondance (assuming no recursive group) known_ids is used to track objects alreay referenced and not include it twice (and possibly missing the fact that the same name can reference multiples different objects) I'm definitively not sure it is correct at all, it is a really really dark corner of Blender :). But it seems to do the job! """ scene_map = blenderapi.get_scene_map()"Scene %s from %s"% (self.scene_name, repr(scene_map.keys()) ) ) self._scene = scene_map[self.scene_name] self._morse_scene = scene_map['S.MORSE_LOGIC'] self._scene_syncable_objects = [] known_ids = set() for obj in self._scene.objects: if != '__default__cam__' and id(obj) not in known_ids: members = obj.groupMembers if not members: self._scene_syncable_objects.append( (obj, self._morse_scene.objects[])) known_ids.add(id(obj)) else: main_members = self._morse_scene.objects[].groupMembers for i in range(0, len(main_members)): self._scene_syncable_objects.append( (members[i], main_members[i])) known_ids.add(id(members[i])) childs = members[i].childrenRecursive main_childs = main_members[i].childrenRecursive for child in childs: self._scene_syncable_objects.append( (child, main_childs[])) known_ids.add(id(child)) def _setup_video_texture(self): """ Prepare this camera to use the bge.texture module. Extract the references to the Blender camera and material where the images will be rendered. """ for child in self.bge_object.children: # The camera object that will produce the image in Blender if 'CameraRobot' in camera = child # The object that contains the material where the image is rendered if 'CameraMesh' in screen = child # Considering it consists of a single mesh mesh = child.meshes[0] # Get the material name for material in mesh.materials: material_index = material.getMaterialIndex() mesh_material_name = mesh.getMaterialName(material_index) if 'MAScreenMat' in mesh_material_name: material_name = mesh_material_name try: logger.debug("\tCAMERA: %s" % logger.debug("\tSCREEN: %s" % logger.debug("\tMATERIAL: %s" % material_name) except UnboundLocalError: logger.error("The video camera could not be properly initialized." "The children object could not be found." "Best solution is to re-link the camera.") return False if not self._offscreen_create: self._compute_syncable_objects() img_renderer = blenderapi.texture().ImageRender(self._scene, camera) else: fbo = self._offscreen_create(self.image_width, self.image_height, blenderapi.render().RAS_OFS_RENDER_TEXTURE) img_renderer = blenderapi.texture().ImageRender(blenderapi.scene(), camera, fbo) mat_id = blenderapi.texture().materialID(screen, material_name) self._camera_image = blenderapi.texture().Texture(screen, mat_id) self._camera_image.source = img_renderer # Set the focal length of the camera using the Game Logic Property. One # can use either focal (lens) or fov parameter: setting one computes the # other accordingly. Fov supersedes focal. if self.image_fov is not None: camera.fov = self.image_fov self.image_focal = camera.lens else: camera.lens = self.image_focal"\tFocal length of the camera is: %s" % camera.lens)"\tFOV of the camera is: %s" % camera.fov) # Set the clipping distances of the camera using the Game Logic Property camera.near = self.near_clipping"\tNear clipping distance of the camera is: %s" % camera.near) camera.far = self.far_clipping"\tFar clipping distance of the camera is: %s" % camera.far) # Set the background to be used for the render self._camera_image.source.background = self.bg_color # Define an image size. It must be powers of two. Default 512 * 512 self._camera_image.source.capsize = [self.image_width, self.image_height]"Camera '%s': Exporting an image of capsize: %s pixels" % (, self._camera_image.source.capsize)) # Workaround capsize limit to window size self.image_width, self.image_height = self._camera_image.source.capsize # Reverse the image (boolean game-property) self._camera_image.source.flip = self.vertical_flip try: # Use the Z-Buffer as an image texture for the camera if self.retrieve_zbuffer: self._camera_image.source.zbuff = True # Use the Z-Buffer as input with an array of depths if self.retrieve_depth: self._camera_image.source.depth = True except AttributeError as detail: logger.warn("%s\nPlease use Blender > 2.65 for Z-Buffer support" % detail)