计算与Intel Realsense R200深度相机的距离

Question

我一直试图用R200摄像机的值来计算物体的距离。我安装了PyRealsense和librealsense(旧式)。PyRealsense的例子没有任何问题。

为此目的，我制定了一项代码：

import pyrealsense as pyrs
from pyrealsense.constants import rs_option
depth_stream = pyrs.stream.DepthStream()
infrared_stream = pyrs.stream.InfraredStream()

with pyrs.Service() as serv:
    with serv.Device(streams=(depth_stream, infrared_stream, )) as dev:
        #dev.apply_ivcam_preset(0)
        while True:
            dev.wait_for_frames()

            print(dev.infrared) 

它返回一个矩阵，该矩阵的值随对象的位置而变化：

 [37 37 39 ... 20 20 21]
 [35 35 38 ... 17 18 19]
 [34 33 37 ... 19 20 20]]
[[40 36 30 ... 16 15 17]
 [40 37 28 ... 14 14 19]
 [42 39 28 ... 14 16 20]

这个矩阵的哪一列表示距离值，或者我应该如何计算距离。

Answer 1

在谷歌上搜索时，我找到了一个用RealSense相机计算距离的例子：

https://github.com/intel/intel-iot-refkit/blob/master/meta-refkit-extra/doc/computervision.rst

为了使它与PyRealSense 2.0一起工作，我不得不编辑它：

#!/usr/bin/python3

import sys

import numpy as np
import cv2
import pyrealsense as pyrs

with pyrs.Service() as serv:
    serv.start()
    with serv.Device() as cam:
        cat_cascade = cv2.CascadeClassifier("/usr/share/opencv/haarcascades/haarcascade_frontalcatface.xml")

        for x in range(30):
            # stabilize exposure
            cam.wait_for_frames()

        while True:
        # get image from web cam
            cam.wait_for_frames()
            img = cam.color

            cats = cat_cascade.detectMultiScale(img)
            for (x,y,w,h) in cats:
                # find center
                cx = int(round(x+(w/2)))
                cy = int(round(y+(h/2)))

                depth = cam.depth[cy][cx]

                print("Cat found, distance " + str(depth/10.0) + " cm")

当它显示猫脸时，它计算距离。我已经开始学习Tensorflow，我对OpenCV的知识很差。您能解释一下吗，将此代码移植到TensorFlow或CAFFE的最简单方法是什么？