如何在列表中保存英特尔实感图像(pyrealsense2)

How to save Intel Realsense images in list (pyrealsense2)

我正在尝试将 Intel Realsense D435i 相机的深度和彩色图像保存在包含 300 张图像的列表中。然后我将使用多处理将这 300 张图像保存到我的磁盘上。但是每次我尝试时,程序都会成功地在列表中附加 15 张图像,然后出现此错误:

    Frame didn't arrived within 5000

我确定我在 python 3.6 上安装了 64 位版本,并且当我不尝试将图像保存在列表中时,相机的流式传输非常好。真实感查看器也很好用。我也尝试过不同的分辨率和帧率,但它似乎也不起作用。有趣的是,如果我只保存彩色图像,我不会得到相同的错误,而是会在列表中一遍又一遍地得到相同的彩色图像。

if __name__ == '__main__':
pipeline = rs.pipeline()
config = rs.config()

config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 30)
config.enable_stream(rs.stream.color, 1280, 720, rs.format.bgr8, 30)
profile = pipeline.start(config)

depth_sensor = profile.get_device().first_depth_sensor()
depth_sensor.set_option(
    rs.option.visual_preset, 3
)  # Set high accuracy for depth sensor
depth_scale = depth_sensor.get_depth_scale()

align_to = rs.stream.color
align = rs.align(align_to)

#   Init variables
im_count = 0
image_chunk = []
image_chunk2 = []
# sentinel = True
try:
    while True:

        # Wait for a coherent pair of frames: depth and color
        frames = pipeline.wait_for_frames()
        aligned_frames = align.process(frames)
        aligned_depth_frame = aligned_frames.get_depth_frame()
        color_frame = aligned_frames.get_color_frame()


        if not aligned_depth_frame or not color_frame:
            print("problem here")
            raise RuntimeError("Could not acquire depth or color frames.")

        depth_image = np.asanyarray(aligned_depth_frame.get_data())
        color_image = np.asanyarray(color_frame.get_data())

        image_chunk.append(color_image)
        image_chunk2.append(depth_image)




except Exception as e:
    print(e)

finally:
    # Stop streaming
    pipeline.stop()

我只是需要它连续保存300张图像,仅此而已,所以我很困惑是什么导致了这个问题。

按住相框会锁定内存,最终会达到极限,从而无法获取更多图像。即使您正在创建图像,数据仍然来自框架。您需要在创建图像后克隆图像以将 link 释放到框架的内存中。

depth_image = np.asanyarray(aligned_depth_frame.get_data())
color_image = np.asanyarray(color_frame.get_data())

depth_image = depth_image.copy()
color_image = color_image.copy()

image_chunk.append(color_image)
image_chunk2.append(depth_image)

在此处阅读有关帧和内存管理的更多信息: https://dev.intelrealsense.com/docs/frame-management

我创建了一个包装器 class 以从框架集中提取以后无法重新创建的各种元素。它有点沉重,但展示了一些可能对其他人有帮助的常见操作:

colorizer = None
align_to_depth = None
align_to_color = None
pointcloud = rs.pointcloud()

class IntelD435ImagePacket:
    """
    Class that contains image and associated processing data.
    """

    @property
    def frame_id(self):
        return self._frame_id

    @property
    def timestamp(self):
        return self._timestamp

    @property
    def image_color(self):
        return self._image_color

    @property
    def image_depth(self):
        return self._image_depth

    @property
    def image_color_aligned(self):
        return self._image_color_aligned

    @property
    def image_depth_aligned(self):
        return self._image_depth_aligned

    @property
    def image_depth_colorized(self):
        if not self._image_depth_colorized:
            self._image_depth_colorized = cv2.applyColorMap(self.image_depth, cv2.COLORMAP_JET);
        return self._image_depth_colorized

    @property
    def intrinsics(self):
        return self._intrinsics

    @property
    def pointcloud(self):
        return self._pointcloud

    @property
    def pointcloud_texture(self):
        return self._pointcloud_texture

    def _rs_intrinsics_to_opencv_matrix(self, rs_intrinsics):
        fx = rs_intrinsics.fx
        fy = rs_intrinsics.fy
        cx = rs_intrinsics.ppx
        cy = rs_intrinsics.ppy
        s = 0  # skew
        return np.array([fx, s, cx,
                         0, fy, cy,
                         0, 0, 1]).reshape(3, 3)

    def __init__(self, frame_set, frame_id=None, timestamp=None, *args, **kwargs):
        global colorizer
        if not colorizer:
            colorizer = rs.colorizer()
            colorizer.set_option(rs.option.color_scheme, 0)

        global align_to_depth
        if not align_to_depth:
            align_to_depth = rs.align(rs.stream.depth)

        global align_to_color
        if not align_to_color:
            align_to_color = rs.align(rs.stream.color)

        global pointcloud
        if not pointcloud:
             pointcloud = rs.pointcloud()

        # Get intrinsics
        profile = frame_set.get_profile()
        video_stream_profile = profile.as_video_stream_profile()
        rs_intrinsics = video_stream_profile.get_intrinsics()
        self._intrinsics = self._rs_intrinsics_to_opencv_matrix(rs_intrinsics)

        # Get pointcloud
        depth_frame = frame_set.get_depth_frame()
        color_frame = frame_set.get_color_frame()
        pointcloud.map_to(color_frame)
        points = pointcloud.calculate(depth_frame)
        vtx = np.asanyarray(points.get_vertices())
        points_arr = vtx.view(np.float32).reshape(vtx.shape + (-1,)).copy()
        self._pointcloud = points_arr

        # Get pointcloud texture mapping
        tex = np.asanyarray(points.get_texture_coordinates())
        color_map_arr = tex.view(np.float32).reshape(tex.shape + (-1,)).copy()
        self._pointcloud_texture = color_map_arr

        # Extract color image
        color_frame = frame_set.get_color_frame()
        self._image_color = np.asanyarray(color_frame.get_data()).copy()

        # Extract depth image
        depth_frame = frame_set.get_depth_frame()
        self._image_depth = np.asanyarray(depth_frame.get_data()).copy()

        # Align the color frame to depth frame and extract color image
        color_frame_aligned = align_to_depth.process(frame_set).get_color_frame()
        self._image_color_aligned = np.asanyarray(color_frame_aligned.get_data()).copy()

        # Align the depth frame to color frame and extract depth image
        depth_frame_aligned = align_to_color.process(frame_set).get_depth_frame()
        self._image_depth_aligned = np.asanyarray(depth_frame_aligned.get_data()).copy()

        self._image_depth_colorized = None
        if frame_id:
            self._frame_id = frame_id
        else:
            self._frame_id = frame_set.frame_number
        if timestamp:
            self._timestamp = timestamp
        else:
            self._timestamp = frame_set.timestamp
        self.__dict__.update(kwargs)