Realsense D435i深度测距和普通摄像头单目测距的区别(附带可用实测代码)

深度摄像头测距

环境：windows+opencv3+python3.64
我采用的是:Realsense D435i 深度摄像头，使用深度摄像头测距的步骤：建一个深度流管、配置流和管道、开启流、创建流对象、对齐流、开启通道后将深度框与颜色框对齐、最后在通过.get_depth_frame()方法获取深度图。获取到深度图后，想要获得深度图上任意一点的距离，即将深度图图像转化为数组、提取点即为对应改点的深度。官方资料仓库

import pyrealsense2 as rs
import numpy as np
import cv2


def nothing(x):
    pass


def creatTrackbar():

    # 蓝灯
    # cv.createTrackbar("hmin", "color_adjust", 0, 255, nothing)
    # cv.createTrackbar("hmax", "color_adjust", 250, 255, nothing)
    # cv.createTrackbar("smin", "color_adjust", 0, 255, nothing)
    # cv.createTrackbar("smax", "color_adjust", 143, 255, nothing)
    # cv.createTrackbar("vmin", "color_adjust", 255, 255, nothing)
    # cv.createTrackbar("vmax", "color_adjust", 255, 255, nothing)
    # 红灯
    cv2.createTrackbar("hmin", "color_adjust", 0, 255, nothing)
    cv2.createTrackbar("hmax", "color_adjust", 30, 255, nothing)
    cv2.createTrackbar("smin", "color_adjust", 5, 255, nothing)
    cv2.createTrackbar("smax", "color_adjust", 100, 255, nothing)
    cv2.createTrackbar("vmin", "color_adjust", 255, 255, nothing)
    cv2.createTrackbar("vmax", "color_adjust", 255, 255, nothing)

    # 形态学操作阈值调整
    cv2.createTrackbar("open", "mor_adjust", 1, 30, nothing)
    cv2.createTrackbar("close", "mor_adjust", 15, 30, nothing)
    cv2.createTrackbar("erode", "mor_adjust", 1, 30, nothing)
    cv2.createTrackbar("dilate", "mor_adjust", 3, 30, nothing)

    # 摄像头调整
    # cv.createTrackbar("gamma", "cap_adjust", 100, 200, nothing)

    cv2.createTrackbar("z", "z_adjust", 100, 360, nothing)


def hsv_change(frame):

    hmin = cv2.getTrackbarPos('hmin', 'color_adjust')
    hmax = cv2.getTrackbarPos('hmax', 'color_adjust')
    smin = cv2.getTrackbarPos('smin', 'color_adjust')
    smax = cv2.getTrackbarPos('smax', 'color_adjust')
    vmin = cv2.getTrackbarPos('vmin', 'color_adjust')
    vmax = cv2.getTrackbarPos('vmax', 'color_adjust')

    # gray = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)
    # cv.imshow("gray", gray)
    hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
    lower_hsv = np.array([hmin, smin, vmin])
    upper_hsv = np.array([hmax, smax, vmax])
    mask = cv2.inRange(hsv, lowerb=lower_hsv, upperb=upper_hsv)
    return mask


# Declare pointcloud object, for calculating pointclouds and texture mappings  声明云对象
pc = rs.pointcloud()
# We want the points object to be persistent so we can display the last cloud when a frame drops
points = rs.points()


pipeline = rs.pipeline()  # 创建一个管道
config = rs.config()  # Create a config并配置要流​​式传输的管道。
config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 15)
config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 15)
# 使用选定的流参数显式启用设备流

# Start streaming 开启流
pipe_profile = pipeline.start(config)

# Create an align object 创建对其流对象
# rs.align allows us to perform alignment of depth frames to others frames
# The "align_to" is the stream type to which we plan to align depth frames.
# (对其流)
align_to = rs.stream.color
align = rs.align(align_to)  # 设置为其他类型的流,意思是我们允许深度流与其他流对齐
print(type(align))
cap = cv2.VideoCapture(0)


def led_practice():
    # creatTrackbar()
    while True:

        frames = pipeline.wait_for_frames()  # 等待开启通道
        # ret, frame = cap.read()  # ret 读取到图片为True 未读到图片为Falst
        # frame = cv2.flip(frame, 1)
        aligned_frames = align.process(frames)  # 将深度框和颜色框对齐
        depth_frame = aligned_frames.get_depth_frame()  # ?获得对齐后的帧数深度数据(图)
        color_frame = aligned_frames.get_color_frame()  # ?获得对齐后的帧数颜色数据(图)
        img_color = np.asanyarray(color_frame.get_data())  # 把图像像素转化为数组
        img_depth = np.asanyarray(depth_frame.get_data())  # 把图像像素转化为数组
        # img_color2 = cv2.cvtColor(img_color, cv2.COLOR_BGR2GRAY)
        # Intrinsics & Extrinsics
        depth_intrin = depth_frame.profile.as_video_stream_profile().intrinsics
        color_intrin = color_frame.profile.as_video_stream_profile().intrinsics
        depth_to_color_extrin = depth_frame.profile.get_extrinsics_to(color_frame.profile)

        # 获取深度传感器的深度标尺
        depth_sensor = pipe_profile.get_device().first_depth_sensor()
        depth_scale = depth_sensor.get_depth_scale()

        # 由深度到颜色
        depth_pixel = [240, 320]  # Random pixel
        depth_point = rs.rs2_deproject_pixel_to_point(depth_intrin, depth_pixel, depth_scale)

        color_point = rs.rs2_transform_point_to_point(depth_to_color_extrin, depth_point)
        color_pixel = rs.rs2_project_point_to_pixel(color_intrin, color_point)
        print('depth: ', color_point)
        print('depth: ', color_pixel)

        pc.map_to(color_frame)
        points = pc.calculate(depth_frame)
        vtx = np.asanyarray(points.get_vertices())  # points.get_vertices() 检索点云的顶点
        tex = np.asanyarray(points.get_texture_coordinates())
        i = 640*200+200
        print('depth: ', [np.float(vtx[i][0]), np.float(vtx[i][1]), np.float(vtx[i][2])])
        cv2.circle(img_color, (300, 250), 8, [255, 0, 255], thickness=-1)
        # cv2.circle(img_color, (300, 250), 8, [255, 0, 255], thickness=-1)

        cv2.putText(img_color, "Distance/cm:"+str(img_depth[300, 250]), (40, 40), cv2.FONT_HERSHEY_SIMPLEX, 1.2, [255, 0, 255])
        cv2.putText(img_color, "X:"+str(np.float(vtx[i][0])), (80, 80), cv2.FONT_HERSHEY_SIMPLEX, 1, [255, 0, 255])
        cv2.putText(img_color, "Y:"+str(np.float(vtx[i][1])), (80, 120), cv2.FONT_HERSHEY_SIMPLEX, 1, [255, 0, 255])
        cv2.putText(img_color, "Z:"+str(np.float(vtx[i][2])), (80, 160), cv2.FONT_HERSHEY_SIMPLEX, 1, [255, 0, 255])
        # cv2.putText api解释:https://blog.csdn.net/weixin_42039090/article/details/80679935
        cv2.imshow('depth_frame', img_color)
        cv2.imshow("dasdsadsa", img_depth)
        # gray = cv2.cvtColor(img_color, cv2.COLOR_BGR2GRAY)

        # cv2.imshow("frame", frames)
        # mask = hsv_change(img_color)
        # cv2.imshow("frame", mask)
        # cv2.imshow('depth_frame', gray)

        key = cv2.waitKey(1)
        if key == 27:
            cv2.destroyAllWindows()
            break
            
led_practice()
cv2.waitKey(0)
cv2.destroyAllWindows()
pipeline.stop()

这里附上我实现的代码，测量中心点的距离。代码注释的都比较详细

单目测距

效果图：


代码：

import numpy as np
import cv2


def nothing(x):
    pass



# 形态学开操作
def open_binary(binary, x, y):
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (x, y))  # 获取图像结构化元素
    dst = cv2.morphologyEx(binary, cv2.MORPH_OPEN, kernel)  # 开操作
    return dst


# 形态学闭操作
def close_binary(binary, x, y):
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (x, y))  # 获取图像结构化元素
    dst = cv2.morphologyEx(binary, cv2.MORPH_CLOSE, kernel)  # 开操作
    return dst


# 形态学腐蚀操作
def erode_binary(binary, x, y):
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (x, y))  # 获取图像结构化元素
    dst = cv2.erode(binary, kernel)  # 腐蚀
    return dst


# 形态学膨胀操作
def dilate_binary(binary, x, y):
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (x, y))  # 获取图像结构化元素
    dst = cv2.dilate(binary, kernel)  # 膨胀返回
    return dst


# 找到目标函数
def find_marker(image):
    img1, contours, heriachy1 = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)  # 获取轮廓点集(坐标)
    cv2.drawContours(frame, contours, -1, (0, 0, 255), 2)
    return contours


# 距离计算函数
def distance_to_camera(knownWidth, focalLength, perWidth):
    # compute and return the distance from the maker to the camera
    if knownWidth * focalLength == 0 or perWidth == 0:
        return 0
    else:
        return (knownWidth * focalLength) / perWidth


def color_detetc(frame):

    hmin1 = cv2.getTrackbarPos('hmin1', 'color_adjust1')
    hmax1 = cv2.getTrackbarPos('hmax1', 'color_adjust1')
    smin1 = cv2.getTrackbarPos('smin1', 'color_adjust1')
    smax1 = cv2.getTrackbarPos('smax1', 'color_adjust1')
    vmin1 = cv2.getTrackbarPos('vmin1', 'color_adjust1')
    vmax1 = cv2.getTrackbarPos('vmax1', 'color_adjust1')
    close = cv2.getTrackbarPos('close', 'mor_adjust')
    erode = cv2.getTrackbarPos('erode', 'mor_adjust')
    dilate = cv2.getTrackbarPos('dilate', 'mor_adjust')
    hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)  # hsv 色彩空间 分割肤色
    lower_hsv1 = np.array([hmin1, smin1, vmin1])
    upper_hsv1 = np.array([hmax1, smax1, vmax1])
    mask1 = cv2.inRange(hsv, lowerb=lower_hsv1, upperb=upper_hsv1)  # hsv 掩码
    ret, thresh1 = cv2.threshold(mask1, 40, 255, cv2.THRESH_BINARY)  # 二值化处理

    dst_close = close_binary(thresh1, close, close)
    dst_erode = erode_binary(dst_close, erode, erode)
    dst_dilate = dilate_binary(dst_erode, dilate, dilate)
    cv2.imshow('dst_close:', dst_close)
    cv2.imshow('dst_erode:', dst_erode)
    cv2.imshow('dst_dilate:', dst_dilate)
    return dst_dilate


if __name__ == "__main__":

    KNOWN_DISTANCE = 24.0
    # A4纸的长和宽(单位:inches)
    KNOWN_WIDTH = 2.69
    KNOWN_HEIGHT = 2.27
    # 通过摄像头标定获取的像素焦距
    focalLength = 811.82
    print('focalLength = ', focalLength)

    # 打开摄像头
    camera = cv2.VideoCapture(0)
    cv2.namedWindow("color_adjust1")
    cv2.namedWindow("mor_adjust")
    cv2.createTrackbar("hmin1", "color_adjust1", 16, 255, nothing)
    cv2.createTrackbar("hmax1", "color_adjust1", 31, 255, nothing)
    cv2.createTrackbar("smin1", "color_adjust1", 119, 255, nothing)
    cv2.createTrackbar("smax1", "color_adjust1", 255, 255, nothing)
    cv2.createTrackbar("vmin1", "color_adjust1", 0, 255, nothing)
    cv2.createTrackbar("vmax1", "color_adjust1", 255, 255, nothing)
    cv2.createTrackbar("close", "mor_adjust", 5, 30, nothing)
    cv2.createTrackbar("erode", "mor_adjust", 10, 30, nothing)
    cv2.createTrackbar("dilate", "mor_adjust", 10, 30, nothing)
    while camera.isOpened():
        # get a frame
        ret, frame = camera.read()
        dst_dilate = color_detetc(frame)

        contours = find_marker(dst_dilate)
        if contours == 0:
            continue
        # compute the bounding box of the of the paper region and return it
        # cv2.minAreaRect() c代表点集，返回rect[0]是最小外接矩形中心点坐标，
        # rect[1][0]是width，rect[1][1]是height，rect[2]是角度
        for i, contour in enumerate(contours):
            area1 = cv2.contourArea(contour)
            if area1 > 100:
                (x1, y1), radius1 = cv2.minEnclosingCircle(contours[i])
                x1 = int(x1)
                y1 = int(y1)
                center1 = (int(x1), int(y1))
                radius1 = int(radius1)
                cv2.circle(frame, center1, 3, (0, 0, 255), -1)  # 画出重心
                print("黄色坐标:", (x1, y1))
                cv2.putText(frame, "yellow:", (x1, y1), cv2.FONT_HERSHEY_SIMPLEX,
                        0.5, [255, 255, 255])
                marker = cv2.minAreaRect(contours[0])
                inches = distance_to_camera(KNOWN_WIDTH, focalLength, marker[1][0])
                # draw a bounding box around the image and display it
                # inches 转换为 cm
                cv2.putText(frame, "%.2fcm" % (inches * 30.48 / 12),
                            (frame.shape[1] - 300, frame.shape[0] - 20), cv2.FONT_HERSHEY_SIMPLEX,
                            2.0, (0, 255, 0), 3)

        # show a frame
        cv2.imshow("capture", frame)
        if cv2.waitKey(1) & 0xFF == 27:
            break
    camera.release()
    cv2.destroyAllWindows()

**说明：**代码为通过hsv识别一个黄色的小球，然后对小球单目测距。摄像头的焦距，是我简单测试的，起个演示作用，最好是通过相机标定测试摄像头的内部参数，然后来实现测距。也可以用几个A4纸测试一下，算出相机焦距。

深度与单目的区别

• 深度测距比较准确，并且不需要所谓目标的实际尺寸，可以随意检测目标距离，而不是特定目标的距离。相机不需要标定知道相机内部的参数。相机标定在出厂时已经标定完了。
• 单目测距精度方面不如深度准确，并且还需要知道相机内部参数，需要相机标定，还得知道测距目标的实际尺寸，只能针对特定目标进行测距。但是优点也是显著的，简单，对相机没有要求。对测距精度不高的环境合适。
• 对比分析：像素和测距的对比。当然用深度的目的远不止用于测距的作用，realsenseD435i的像素不算很高，即想看的清，又想测得准的话，realsenseD435i是不太实用的。单目虽然测得不算很准，但是相比之下可以选择像素高的单目相机，同样还是可以测距。但是realsenseD435i测距准确,但是像素固定无法提高了。看适用场合选取合适的摄像头。