自动驾驶采集多视角图像处理（python实现不同文件夹下同名图片批量拼接并生成视频）

目录

1.图像拼接

2.合成视频

2.1 cv2.videowriter_fourcc功能介绍

2.2读取图片合成视频

1.图像拼接

实现目标：将自动驾驶车6个摄像头采集到的图片，按照正确顺序拼接显示，nuScenes数据集测试如下：

图像存放文件夹目录如下：

每个目录下有相同名称，不同视角采集到的图像 。

先显示图片，代码如下：

import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt
import os



"""
将六张图片在同一窗口下显示，关闭后打开后一帧的六张拼接图片
"""

def show_images(images, titles=None, num_cols=None, scale=3, normalize=False):
    """ 一个窗口中绘制多张图像:
    Args:
        images: 可以为一张图像(不要放在列表中)，也可以为一个图像列表
        titles: 图像对应标题、
        num_cols: 每行最多显示多少张图像
        scale: 用于调整图窗大小
        normalize: 显示灰度图时是否进行灰度归一化
    """

    # 加了下面2行后可以显示中文标题
    plt.rcParams['font.sans-serif'] = ['SimHei']
    plt.rcParams['axes.unicode_minus'] = False
    """
    # 单张图片显示
    if not isinstance(images, list):
        if not isinstance(scale, tuple):
            scale = (scale, scale * 1.5)

        plt.figure(figsize=(scale[1], scale[0]))
        img = images
        if len(img.shape) == 3:
            # opencv库中函数生成的图像为BGR通道，需要转换一下
            B, G, R = cv.split(img)
            img = cv.merge([R, G, B])
            plt.imshow(img)
        elif len(img.shape) == 2:
            # pyplot显示灰度需要加一个参数
            if normalize:
                plt.imshow(img, cmap='gray')
            else:
                plt.imshow(img, cmap='gray', vmin=0, vmax=255)
        else:
            raise TypeError("Invalid shape " +
                            str(img.shape) + " of image data")
        if titles is not None:
            plt.title(titles, y=-0.15)
        plt.axis('off')
        plt.show()
        return
    """
    # 多张图片显示
    if not isinstance(scale, tuple):
        scale = (scale, scale)
    num_imgs = len(images)
    if num_cols is None:
        num_cols = int(np.ceil((np.sqrt(num_imgs))))
    num_rows = (num_imgs - 1) // num_cols + 1

    idx = list(range(num_imgs))
    _, figs = plt.subplots(num_rows, num_cols,
                           figsize=(scale[1] * num_cols, scale[0] * num_rows))

    for f, i, img in zip(figs.flat, idx, images):
        if len(img.shape) == 3:
            # opencv库中函数生成的图像为BGR通道，需要转换一下
            B, G, R = cv.split(img)
            img = cv.merge([R, G, B])
            f.imshow(img)
        elif len(img.shape) == 2:
            # pyplot显示灰度需要加一个参数
            if normalize:
                f.imshow(img, cmap='gray')
            else:
                f.imshow(img, cmap='gray', vmin=0, vmax=255)
        else:
            raise TypeError("Invalid shape " +
                            str(img.shape) + " of image data")
        if titles is not None:
            f.set_title(titles[i], y=-0.15)
        f.axes.get_xaxis().set_visible(False)
        f.axes.get_yaxis().set_visible(False)
    # 将不显示图像的fig移除，不然会显示多余的窗口
    if len(figs.shape) == 1:
        figs = figs.reshape(-1, figs.shape[0])
    for i in range(num_rows * num_cols - num_imgs):
        figs[num_rows - 1, num_imgs % num_cols + i].remove()
    plt.show()


if __name__ == '__main__':
    """
    src1= cv.imread("vis_result/camera-2")
    src2 = cv.imread("vis_result/camera-0")
    src3 = cv.imread("vis_result/camera-1")
    src4 = cv.imread("vis_result/camera-4")
    src5 = cv.imread("vis_result/camera-3")
    src6 = cv.imread("vis_result/camera-5")
    """
    dir0 = "D:/pythonproject/bevshow/vis_result"
    dir1 = r"D:/pythonproject/bevshow/vis_result/camera-0/"
    path = os.listdir(dir1)
    for name in path:
        src1 = cv.imread(os.path.join(dir0, "camera-2", name))
        src2 = cv.imread(os.path.join(dir0, "camera-0", name))
        src3 = cv.imread(os.path.join(dir0, "camera-1", name))
        src4 = cv.imread(os.path.join(dir0, "camera-5", name))
        src5 = cv.imread(os.path.join(dir0, "camera-3", name))
        src6 = cv.imread(os.path.join(dir0, "camera-4", name))
        title1 = "front_left"
        title2 = "front"
        title3 = "front_right"
        title4 = "back_right"
        title5 = "back"
        title6 = "back_left"
        show_images([src1, src2, src3, src4, src5, src6], [title1, title2, title3, title4, title5, title6], num_cols=3)

    # show_images(src, title)

实现效果为将6个文件夹中对应帧图片（同名图片）在一个窗口中显示，并显示标题：

拼接图像并保存，实现参考这篇：

使用Python批量拼接图片

代码如下：

import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt
import os
from PIL import Image

"""
输入六张图片文件夹，输出逐帧拼接的图片并保存在指定目录下
"""
# 保存路径
image_dir = 'D:/pythonproject/bevshow/result/'
# 6个文件夹根目录
dir0 = "D:/pythonproject/bevshow/vis_result"
# 其中一个文件夹，读取图片名用
dir1 = r"D:/pythonproject/bevshow/vis_result/camera-0/"

if __name__ == '__main__':
    """
    src1= cv.imread("vis_result/camera-2")
    src2 = cv.imread("vis_result/camera-0")
    src3 = cv.imread("vis_result/camera-1")
    src4 = cv.imread("vis_result/camera-4")
    src5 = cv.imread("vis_result/camera-3")
    src6 = cv.imread("vis_result/camera-5")
    """

    path = os.listdir(dir1)
    for name in path:
        images = []
        images0 = Image.open(os.path.join(dir0, "camera-2", name))
        # print(os.path.join(dir0, "camera-2", name))
        # D:/pythonproject/bevshow/vis_result\camera-2\1533151603547590-3e8750f331d7499e9b5123e9eb70f2e2.png
        images1 = Image.open(os.path.join(dir0, "camera-0", name))
        images2 = Image.open(os.path.join(dir0, "camera-1", name))
        images3 = Image.open(os.path.join(dir0, "camera-4", name))
        images4 = Image.open(os.path.join(dir0, "camera-3", name))
        images5 = Image.open(os.path.join(dir0, "camera-5", name))
        # 如果大小不同可以resize一下
        """
        images0 = images0.resize((1600, 900))
        images1 = images1.resize((1600, 900))
        images2 = images2.resize((1600, 900))
        images3 = images3.resize((1600, 900))
        images4 = images4.resize((1600, 900))
        images5 = images5.resize((1600, 900))
        """
        # 拼接图片，按照自己想要的顺序，从左到右，从上到下
        images.append(images0)
        images.append(images1)
        images.append(images2)
        images.append(images3)
        images.append(images4)
        images.append(images5)

        new_image = Image.new('RGB', (4800, 1800), 'white')

        # 将四个图像粘贴到新图像的正确位置
        new_image.paste(images[0], (0, 0))
        new_image.paste(images[1], (1600, 0))
        new_image.paste(images[2], (3200, 0))
        new_image.paste(images[3], (0, 900))
        new_image.paste(images[4], (1600, 900))
        new_image.paste(images[5], (3200, 900))
        # 将最终图像保存到磁盘上
        new_image.save(image_dir + name)
        print(new_image)
        # show_images([src1, src2, src3, src4, src5, src6], [title1, title2, title3, title4, title5, title6], num_cols=3)

    # show_images(src, title)

将拼接好的图像会保存在提前建立的result文件夹中

2.合成视频

读取result文件夹中拼接好的图像，按照一定帧数合成为视频

参考：python 将图片合成视频

2.1 cv2.videowriter_fourcc功能介绍

  cv2.videowriter_fourcc是OpenCV中的一个函数，用于创建一个视频编码器对象。它接受四个字符作为参数，用于指定视频编码器的类型。这些字符通常是四个大写字母，例如MJPG、XVID、DIVX等。使用cv2.videowriter_fourcc可以将视频数据编码为指定类型的视频文件，以便在其他设备上播放或编辑。

参数列表
cv2.VideoWriter_fourcc('M', 'P', '4', 'V')
MPEG-4编码 .mp4 可指定结果视频的大小
cv2.VideoWriter_fourcc('X','2','6','4')
MPEG-4编码 .mp4 可指定结果视频的大小
cv2.VideoWriter_fourcc('I', '4', '2', '0')
该参数是YUV编码类型，文件名后缀为.avi 广泛兼容，但会产生大文件
cv2.VideoWriter_fourcc('P', 'I', 'M', 'I')
该参数是MPEG-1编码类型，文件名后缀为.avi
cv2.VideoWriter_fourcc('X', 'V', 'I', 'D')
该参数是MPEG-4编码类型，文件名后缀为.avi，可指定结果视频的大小
cv2.VideoWriter_fourcc('T', 'H', 'E', 'O')
该参数是Ogg Vorbis,文件名后缀为.ogv
cv2.VideoWriter_fourcc('F', 'L', 'V', '1')
该参数是Flash视频，文件名后缀为.flv

报错1：

[ WARN:0@0.033] global loadsave.cpp:248 cv::findDecoder imread_('D:/pythonproject/bevshow/result1533151603547590-3e8750f331d7499e9b5123e9eb70f2e2.png'): can't open/read file: check file path/integrity
Traceback (most recent call last):
  File "D:\pythonproject\bevshow\test03.py", line 15, in <module>
    img = cv2.resize(img, (1280, 720))  # 将图片转换为1280*720
cv2.error: OpenCV(4.8.1) D:\a\opencv-python\opencv-python\opencv\modules\imgproc\src\resize.cpp:4062: error: (-215:Assertion failed) !ssize.empty() in function 'cv::resize'

解决方案：查看图片是否为空，若为空则是路径问题，修改后即可

2.2读取图片合成视频

此处读取的图片格式为png，参考这一篇：【Python入门教程】基于OpenCV视频分解成图片+图片组合成视频（视频抽帧组帧）

代码如下：

# -*- coding: utf-8 -*-
"""
图片组帧成视频
"""
import os
import cv2
from PIL import Image


def Image_Frame(images_path, out_path, fps):
    """
    :param images_path: 输入需要组帧的图片文件夹路径
    :param out_path: 输出视频路径
    :param fps: 视频帧率
    :return: None
    """
    print("-------------------------图片组帧-------------------------")
    images_lists = os.listdir(images_path)  # images_lists.sort()
    image_size = Image.open(os.path.join(images_path, images_lists[0])).size
    fourcc = cv2.VideoWriter_fourcc(*"mp4v")
    video_writer = cv2.VideoWriter(out_path, fourcc, fps, image_size)
    for image_list in images_lists:
        image_path = os.path.join(images_path, image_list)
        frame = cv2.imread(image_path)
        video_writer.write(frame)
        print("正在添加:", image_list)
    video_writer.release()
    print("-------------------------组帧完成-------------------------")


if __name__ == "__main__":
    Image_path = r'D:\pythonproject\bevshow\vis_result\camera-0'
    Out_path = r'D:/pythonproject/bevshow/testdir/1.mp4'
    FPS = int(input("请输入帧率："))
    Image_Frame(Image_path, Out_path, FPS)

在运行时一切都正常，但就是文件夹里没有合成的视频，我也不清楚什么原因，后来问了作者，发现是out_path要到文件名，不能只到文件夹，修改后可以生成而且速度比较快。

换了里另一个代码试了一下：python学习之合并多张图片转成mp4视频代码实现

 报错2：

ValueError: append_data requires ndarray as first arg

解决：

        frame = image.resize((1600, 912))
        frame = np.array(frame)

运行成功代码为：

import os
import imageio
import numpy as np
from PIL import Image

"""
读取拼接好的图片，将其生成指定帧数视频并保存在指定目录下
"""
# 设置生成的视频文件名和路径
filename = 'output.mp4'#文件名
filepath = os.path.join("D:/pythonproject/bevshow/testdir", filename)#存储路径
path = "D:/pythonproject/bevshow/result"#图片读取路径

# 读取所有 PNG 图片
images = []
for file_name in sorted(os.listdir(path)):
    if file_name.endswith('.png'):
        img = Image.open(os.path.join(path, file_name))
        images.append(img)

# 将图片转换为视频
fps = 30  # 每秒钟30帧
with imageio.get_writer(filepath, fps=fps) as video:
    for image in images:
        frame = image.resize((1600, 912))
        frame = np.array(frame)
        video.append_data(frame)

在指定目录下生成了MP4格式视频：

可以调整帧数以及尺寸大小，调尺寸可能会报错，但是可以在报错里找到给出的合适尺寸，照着修改即可 ，视频效果如下：

3.完整代码

import os
import imageio
import numpy as np
from PIL import Image

"""
将两个步骤整合，输入6个文件夹，直接输出视频
"""

"""
输入六张图片文件夹，输出逐帧拼接的图片并保存在指定目录下
"""
# 保存路径
image_dir = 'D:/pythonproject/bevshow/result/'
# 6个文件夹根目录
dir0 = "D:/pythonproject/bevshow/vis_result"
# 其中一个文件夹，读取图片名用
dir1 = r"D:/pythonproject/bevshow/vis_result/camera-0/"
filename = 'output1.mp4'  # 保存视频名称
filepath = os.path.join("D:/pythonproject/bevshow/testdir", filename)  # 保存视频路径

if __name__ == '__main__':
    path = os.listdir(dir1)
    img = []
    for name in path:
        images = []
        images0 = Image.open(os.path.join(dir0, "camera-2", name))  # 左前
        images1 = Image.open(os.path.join(dir0, "camera-0", name))  # 前
        images2 = Image.open(os.path.join(dir0, "camera-1", name))  # 右前
        images3 = Image.open(os.path.join(dir0, "camera-5", name))  # 右后
        images4 = Image.open(os.path.join(dir0, "camera-3", name))  # 后
        images5 = Image.open(os.path.join(dir0, "camera-4", name))  # 左后
        # 拼接图片，按照自己想要的顺序，从左到右，从上到下
        images.append(images0)
        images.append(images1)
        images.append(images2)
        images.append(images3)
        images.append(images4)
        images.append(images5)

        new_image = Image.new('RGB', (4800, 1800), 'white')

        # 将四个图像粘贴到新图像的正确位置
        new_image.paste(images[0], (0, 0))
        new_image.paste(images[1], (1600, 0))
        new_image.paste(images[2], (3200, 0))
        new_image.paste(images[3], (0, 900))
        new_image.paste(images[4], (1600, 900))
        new_image.paste(images[5], (3200, 900))
        img.append(new_image)


    # 将图片转换为视频
    fps = 10  # 每秒钟30帧
    with imageio.get_writer(filepath, fps=fps) as video:
        for image in img:
            frame = image.resize((1872, 1088))
            frame = np.array(frame)
            video.append_data(frame)

到了这里，关于自动驾驶采集多视角图像处理（python实现不同文件夹下同名图片批量拼接并生成视频）的文章就介绍完了。如果您还想了解更多内容，请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章，希望大家以后多多支持TOY模板网！