我正在尝试为 MSRKINECTNUI.DLL 中公开的 MSoft Kinect 函数编写一个包装器。
我正在转换官方驱动程序附带的骨架查看器示例。此示例使用 C++ 并使用 DLL 中的相同函数。可以在此处阅读描述此示例的原始 C++ 文档。
你可以在我的github上看到进度。转子工作，事件似乎发生了。但这是很多黑客行为。现在我有点卡住了。
这是在顶部定义了 ctype 结构的 kinect 接口类。哦，是的，你需要一个 Kinect，在带有 Microsoft 的 Kinect Beta 驱动程序的 Windows 上运行。
# PYTHON 2.7
import sys,os
import time
import thread
import atexit
from datetime import datetime
import Image
#from pyglet import app
#from pyglet import window
from ctypes import *
from ctypes.wintypes import HANDLE, ULONG, DWORD, BOOL, LPCSTR, LPCWSTR, WinError, HRESULT, WORD, BYTE
from ctypes.util import find_library
import win32event
from win32process import beginthreadex, GetExitCodeProcess
from win32api import GetCurrentThreadId
from win32ui import CreateThread
from threading import Thread
"""
version 0.01
Please check Changes.txt for changes and the development planning
"""
# NUI Common Initialization Declarations
NUI_INITIALIZE_FLAG_USES_DEPTH_AND_PLAYER_INDEX = 0x00000001
NUI_INITIALIZE_FLAG_USES_COLOR = 0x00000002
NUI_INITIALIZE_FLAG_USES_SKELETON = 0x00000008
NUI_INITIALIZE_FLAG_USES_DEPTH = 0x00000020
NUI_INITIALIZE_DEFAULT_HARDWARE_THREAD = 0xFFFFFFFF
# _NUI_IMAGE_TYPE
NUI_IMAGE_TYPE_DEPTH_AND_PLAYER_INDEX = 0 #/ USHORT
NUI_IMAGE_TYPE_COLOR = 1 # RGB32 data
NUI_IMAGE_TYPE_COLOR_YUV = 2 # YUY2 stream from camera h/w but converted to RGB32 before user getting it.
NUI_IMAGE_TYPE_COLOR_RAW_YUV = 3 # YUY2 stream from camera h/w.
NUI_IMAGE_TYPE_DEPTH = 4 # USHORT
NUI_IMAGE_TYPE_DEPTH_AND_PLAYER_INDEX_IN_COLOR_SPACE = 5 # not available yet
NUI_IMAGE_TYPE_DEPTH_IN_COLOR_SPACE = 6 # not available yet
NUI_IMAGE_TYPE_COLOR_IN_DEPTH_SPACE = 7
# _NUI_IMAGE_RESOLUTION
NUI_IMAGE_RESOLUTION_INVALID =-1
NUI_IMAGE_RESOLUTION_80x60 = 0
NUI_IMAGE_RESOLUTION_320x240 = 1
NUI_IMAGE_RESOLUTION_640x480 = 2
NUI_IMAGE_RESOLUTION_1280x1024 = 3 # for hires color only
NUI_IMAGE_PLAYER_INDEX_SHIFT = 3
NUI_IMAGE_PLAYER_INDEX_MASK = ((1 << NUI_IMAGE_PLAYER_INDEX_SHIFT)-1)
NUI_IMAGE_DEPTH_MAXIMUM = ((4000 << NUI_IMAGE_PLAYER_INDEX_SHIFT) | NUI_IMAGE_PLAYER_INDEX_MASK)
NUI_IMAGE_DEPTH_MINIMUM = (800 << NUI_IMAGE_PLAYER_INDEX_SHIFT)
NUI_IMAGE_DEPTH_NO_VALUE = 0
NUI_CAMERA_DEPTH_NOMINAL_FOCAL_LENGTH_IN_PIXELS = 285.63 # Based on 320x240 pixel size.
NUI_CAMERA_DEPTH_NOMINAL_INVERSE_FOCAL_LENGTH_IN_PIXELS = 3.501e-3 # (1/NUI_CAMERA_DEPTH_NOMINAL_FOCAL_LENGTH_IN_PIXELS)
NUI_CAMERA_DEPTH_NOMINAL_DIAGONAL_FOV = 70.0
NUI_CAMERA_DEPTH_NOMINAL_HORIZONTAL_FOV = 58.5
NUI_CAMERA_DEPTH_NOMINAL_VERTICAL_FOV = 45.6
NUI_CAMERA_COLOR_NOMINAL_FOCAL_LENGTH_IN_PIXELS = 531.15 # Based on 640x480 pixel size.
NUI_CAMERA_COLOR_NOMINAL_INVERSE_FOCAL_LENGTH_IN_PIXELS = 1.83e-3 # (1/NUI_CAMERA_COLOR_NOMINAL_FOCAL_LENGTH_IN_PIXELS)
NUI_CAMERA_COLOR_NOMINAL_DIAGONAL_FOV = 73.9
NUI_CAMERA_COLOR_NOMINAL_HORIZONTAL_FOV = 62.0
NUI_CAMERA_COLOR_NOMINAL_VERTICAL_FOV = 48.6
# the max # of NUI output frames you can hold w/o releasing
NUI_IMAGE_STREAM_FRAME_LIMIT_MAXIMUM = 4
# return S_FALSE instead of E_NUI_FRAME_NO_DATA if # NuiImageStreamGetNextFrame( )
# doesn't have a frame ready and a timeout != INFINITE is used
NUI_IMAGE_STREAM_FLAG_SUPPRESS_NO_FRAME_DATA = 0x00010000
# Camera extreme angles
Camera_ElevationMaximum = 25 # actual 27 till -27
Camera_ElevationMinimum =-25
# others
S_OK = 0
#HANDLE = HRESULT
####################
# using ctypes to talk to kinect drivers
# http://docs.python.org/library/ctypes.html
# http://python.net/crew/theller/ctypes/tutorial.html#calling-functions
# create event handlers
m_hEvNuiProcessStop = win32event.CreateEvent(None, 0, 0, None)
m_hNextDepthFrameEvent = win32event.CreateEvent(None, 1, 0, None)
m_hNextVideoFrameEvent = win32event.CreateEvent(None, 1, 0, None)
m_hNextSkeletonEvent = win32event.CreateEvent(None, 1, 0, None)
hEvents = [ m_hEvNuiProcessStop,
m_hNextDepthFrameEvent,
m_hNextVideoFrameEvent,
m_hNextSkeletonEvent ]
#m_pVideoStreamHandle = pointer(c_int(1))
#m_pDepthStreamHandle = pointer(c_int(1))
m_pVideoStreamHandle = c_void_p()
m_pDepthStreamHandle = c_void_p()
class _NUI_IMAGE_VIEW_AREA(Structure):
_fields_ = [("eDigitalZoom_NotUsed",c_int),
("lCenterX_NotUsed",c_long),
("lCenterY_NotUsed",c_long)]
class _NuiImageBuffer(Structure):
_fields_= [("m_Width",c_int),
("m_Height",c_int),
("m_BytesPerPixel",c_int),
("m_pBuffer",c_void_p),
("BufferLen",c_int), # m_Width * m_Height * m_BytesPerPixel;
("Pitch",c_int)] # m_Width * m_BytesPerPixel;
class _NUI_IMAGE_FRAME(Structure):
_fields_ = [("liTimeStamp", c_longlong),
("dwFrameNumber", c_ulong),
("eImageType", c_int), # enum NUI_IMAGE_TYPE
("eResolution", c_int), # enum NUI_IMAGE_RESOLUTION
("pFrameTexture", POINTER(_NuiImageBuffer)), # obj from class NuiImageBuffer
("dwFrameFlags_NotUsed", c_ulong),
("ViewArea_NotUsed", _NUI_IMAGE_VIEW_AREA)]
pImageFrame = _NUI_IMAGE_FRAME()
class KinectInterface:
# dll files
dll_Kinect = "MSRKINECTNUI.DLL"
lock = thread.allocate_lock()
nrKinects = 0
def __init__(self):
if 'win32' != sys.platform:
raise Exception('Only supports x86 windows.')
if find_library(self.dll_Kinect):
try:
self.dll = windll.LoadLibrary(self.dll_Kinect)
except:
raise Exception('Kinect drivers installed but can not be loaded')
else:
raise Exception('Kinect drivers not installed.')
# self.m_hThNuiProcess = Nui_ProcessThread("m_hThNuiProcess", self)
# check how many kinects are connected to the system
self.nrKinects = self.nrKinectsConnected()
atexit.register(self.goodbye)
if self.nrKinects == 0:
raise Exception('No kinect connected')
def Nui_ProcessThread(self,*args):
print "Nui_ProcessThread started:",args
# http://stackoverflow.com/questions/100624/python-on-windows-how-to-wait-for-multiple-child-processes
abort=0
while not abort:
nEventIdx = win32event.WaitForMultipleObjects(hEvents, False, 100)
if nEventIdx == 0:
print "m_hEvNuiProcessStop event occurred:",hEvents[nEventIdx]
abort = 1
elif nEventIdx == 1:
print "m_hNextDepthFrameEvent event occurred:",hEvents[nEventIdx]
elif nEventIdx == 2:
print "m_hNextVideoFrameEvent event occurred:",hEvents[nEventIdx]
self.Nui_GotVideoAlert()
elif nEventIdx == 3:
print "m_hNextSkeletonEvent event occurred:",hEvents[nEventIdx]
self.Nui_GotVideoAlert()
else:
print "unknown nEventIdx:",nEventIdx
#finished = handles[nEventIdx]
#exitcode = GetExitCodeProcess(finished)
#procname = hEvents.pop(finished)
#finished.close()
#print "Subprocess %s finished with exit code %d" % (procname, exitcode)
time.sleep(0.3)
print "Nui_ProcessThread ended"
def getImageFrame(self):
hr=cdll.MSRKINECTNUI.NuiImageGetNextFrame()
if hr != S_OK:
print "getImageFrame failed:",hr
def getSkeletonFrame(self):
hr=cdll.MSRKINECTNUI.NuiSkeletonGetNextFrame()
if hr != S_OK:
print "getSkeletonFrame failed:",hr
def Nui_Init(self):
# todo: use pyglet to handle video window
self.NuiInitialize()
#self.NuiSkeletonTrackingEnable()
self.NuiImageVideoStreamOpen()
#self.NuiImageDepthStreamOpen()
# create the worker threads
ProcessThread=Thread(target=self.Nui_ProcessThread, args=())
ProcessThread.setDaemon(True)
ProcessThread.start()
while ProcessThread.isAlive():
time.sleep(1)
def NuiInitialize(self,options = None):
if options == None:
options = NUI_INITIALIZE_FLAG_USES_COLOR | \
NUI_INITIALIZE_FLAG_USES_DEPTH_AND_PLAYER_INDEX | \
NUI_INITIALIZE_FLAG_USES_SKELETON
hr=cdll.MSRKINECTNUI.NuiInitialize( options )
if hr != S_OK:
raise Exception('NuiInitialize failed.')
def NuiSkeletonTrackingEnable(self):
print "NuiSkeletonTrackingEnable"
cdll.MSRKINECTNUI.NuiSkeletonTrackingEnable.restype= c_uint
cdll.MSRKINECTNUI.NuiSkeletonTrackingEnable.argtypes = [HRESULT, c_uint]
hr = cdll.MSRKINECTNUI.NuiSkeletonTrackingEnable( m_hNextSkeletonEvent, c_uint(0))
if hr != S_OK:
raise Exception('NuiSkeletonTrackingEnable failed.')
def NuiImageVideoStreamOpen(self):
print "NuiImageVideoStreamOpen"
cdll.MSRKINECTNUI.NuiImageStreamOpen.argtypes = [c_int, c_int,c_int,c_int, HRESULT , c_void_p]
cdll.MSRKINECTNUI.NuiImageStreamOpen.restype = c_uint
hr = cdll.MSRKINECTNUI.NuiImageStreamOpen(NUI_IMAGE_TYPE_COLOR,
NUI_IMAGE_RESOLUTION_640x480,
0,2,
m_hNextVideoFrameEvent,
byref(m_pVideoStreamHandle))
if hr != S_OK:
raise Exception('NuiImageStreamOpen (NuiImageVideoStreamOpen) failed.')
def NuiImageDepthStreamOpen(self):
print "NuiImageDepthStreamOpen"
cdll.MSRKINECTNUI.NuiImageStreamOpen.argtypes = [c_uint, c_uint,c_uint,c_uint, HRESULT , c_void_p]
cdll.MSRKINECTNUI.NuiImageStreamOpen.restype = c_uint
hr = cdll.MSRKINECTNUI.NuiImageStreamOpen(NUI_IMAGE_TYPE_DEPTH_AND_PLAYER_INDEX,
NUI_IMAGE_RESOLUTION_320x240,
0,2,
m_hNextDepthFrameEvent,
byref(_pDepthStreamHandle))
if hr != S_OK:
raise Exception('NuiImageStreamOpen (NuiImageDepthStreamOpen) failed.')
def nrKinectsConnected(self):
nrKinects = c_int(0)
hr=cdll.MSRKINECTNUI.MSR_NUIGetDeviceCount(byref(nrKinects))
if hr != S_OK:
raise Exception('error white getting nr of Kinects.')
print "Kinects connected:",nrKinects.value
return nrKinects.value
def Nui_GotVideoAlert(self):
pImageFrame.values = None
# reserve buffer
#pImageFrame.pFrameTexture = c_ubyte * (640*480)
#pImageFrame.pFrameTexture = pImageFrame.pFrameTexture()
cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame.argtypes = [c_void_p, c_uint, POINTER(_NUI_IMAGE_FRAME)]
cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame.restype = c_uint
hr=cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame(m_pVideoStreamHandle,0,byref(pImageFrame));
#cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame.argtypes = [c_void_p, c_uint, _NUI_IMAGE_FRAME]
#cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame.restype = c_uint
#hr=cdll.MSRKINECTNUI.NuiImageStreamGetNextFrame(m_pVideoStreamHandle,0,pImageFrame);
if hr != S_OK:
raise Exception('Nui_GotVideoAlert failed.')
print "timeStamp:",pImageFrame.liTimeStamp,datetime.fromtimestamp(pImageFrame.liTimeStamp).strftime("%d/%m/%y %H:%M")
#pImage = pImageFrame.pFrameTexture
#INTP = POINTER(c_long)
#print "pImage:",type(pImage)
#addr = addressof(pImage)
#print ' address: %x'%(addr), type(addr)
#ptr = cast(addr, INTP)
#print ' pointer:', ptr
#print ' value:', ptr[0]
#graydata = cast(graydata, POINTER(c_ubyte))
#self.convertImageFromString(imgSize,buffer)
#mgBuffer=pImage.values
#imgBuffer = c_ubyte * (640*480)()
#cast(imgBuffer, POINTER(c_int))
#cast(pImage,)
#imgBuffer = cast(pImage, POINTER(c_ubyte))
#print "imgBuffer:",type(imgBuffer)
#imgSize = (int(640), int(480))
#self.convertImage(imgSize,pImage)
cdll.MSRKINECTNUI.NuiImageStreamReleaseFrame( m_pVideoStreamHandle, pImageFrame );
# KINECT_LOCKED_RECT LockedRect;
# pTexture->LockRect( 0, &LockedRect, NULL, 0 );
# if( LockedRect.Pitch != 0 )
# {
# BYTE * pBuffer = (BYTE*) LockedRect.pBits;
# m_DrawVideo.DrawFrame( (BYTE*) pBuffer );
# }
# else
# {
# OutputDebugString( L"Buffer length of received texture is bogus\r\n" );
# }
def convertImageFromString(self,imgSize,imgString):
im = Image.fromstring('L', imgSize, imgString)
im.save('out.png')
def convertImage(self,imgSize,bytes):
im = Image.frombuffer('L', (10,10), bytes,'raw', 'L', 0, 1)
#im = Image.fromarray('L', bytes)
#im = Image.fromstring('L', imgSize, bytes) #, 'raw', 'F;16')
im.save('out.png')
#PlanarImage Image = e.ImageFrame.Image;
#video.Source = BitmapSource.Create(
#Image.Width, Image.Height, 96, 96, PixelFormats.Bgr32, null,
#Image.Bits, Image.Width * Image.BytesPerPixel)
def __del__(self):
print "Shutting down kinect"
cdll.MSRKINECTNUI.NuiShutdown()
#windll.kernel32.CloseHandle(lpProcessInformation.hProcess)
def goodbye(self):
print "Application ended."
def getAngle(self):
angle = c_int(0)
cdll.MSRKINECTNUI.NuiCameraElevationGetAngle(byref(angle))
return angle.value
def setAngle(self,angle=0):
print "moving Kinect to angle",angle
if self.getAngle() == angle:
return
angle=max(angle,Camera_ElevationMinimum) # clamp the angle
angle=min(angle,Camera_ElevationMaximum) # bewteen min and max
hr = cdll.MSRKINECTNUI.NuiCameraElevationSetAngle(c_int(angle));
pa = self.getAngle()
time.sleep(1)
while (self.getAngle() != pa):
time.sleep(0.1)
pa = self.getAngle()
def testAll(self):
print("== Test motor")
print(" moving to 0 degrees")
myKinect.setAngle(0)
print(" moving to -20 degrees")
myKinect.setAngle(-20)
print(" moving to 20 degrees")
myKinect.setAngle(20)
print(" moving to 0 degrees")
myKinect.setAngle(0)
print("== Test motor ended")
# simple unit test
if __name__ == '__main__':
myKinect = KinectInterface()
if myKinect.nrKinects == 0:
print("exiting.")
os._exit(-1)
myKinect.Nui_Init()
del myKinect
os._exit(0)
代码很乱，正在进行中。我故意留下评论，试图表明我是多么绝望（或懒惰:-)。无论如何，问题是 pImageFrame 对象没有被 NuiImageStreamGetNextFrame() 填充。我通过没有更新的时间戳得出结论。下一个问题是访问 pFrameTexture BYTE 数据并将其显示为 PIL 中的图像。但那是下一章。
输出：
Kinects 已连接：1 NuiImageVideoStreamOpen Nui_ProcessThread 已启动：() m_hNextVideoFrameEvent 事件发生：timeStamp：82807608 16/08/72 18:06 m_hNextVideoFrameEvent 事件发生：timeStamp：82807608 16/08/72 18:06 m_hNextVideoFrameEvent 事件发生：timeStamp：82807608 16/08 /72 18:06 m_hNextVideoFrameEvent 事件发生：timeStamp：82807608 16/08/72 18:06 m_hNextVideoFrameEvent 事件发生：timeStamp：82807608 16/08/72 18:06