python opencv 读取mp4,上一帧，下一帧

目录

python opencv 读取mp4, 有上一帧，下一帧的功能

python 读取mp4,缓存图片在队列中，这个速度快

队列封装成类版本：

---

python opencv 读取mp4, 有上一帧，下一帧的功能

这个每次读取self.cap.set(cv2.CAP_PROP_POS_FRAMES, frame_index)，后退读取稍微有点慢，

import cv2
from pynput import keyboard

class VideoPlayer:
    def __init__(self, video_path):
        self.cap = cv2.VideoCapture(video_path)
        self.frame_index = -1
        self.total_frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))
        self.new_frame_flag = False
        self.last_frame_i=0

    def show_frame(self, frame_index):
        if 0 <= frame_index < self.total_frames:
            if frame_index<self.last_frame_i:
                self.cap.set(cv2.CAP_PROP_POS_FRAMES, frame_index)
            self.last_frame_i = frame_index
            ret, frame = self.cap.read()
            if ret:
                cv2.imshow('Video', frame)
                self.frame_index = frame_index
                self.new_frame_flag = False

    def next_frame(self):
        if self.frame_index + 1 < self.total_frames:
            self.frame_index += 1
            self.new_frame_flag = True

    def prev_frame(self):
        if self.frame_index - 1 >= 0:
            self.frame_index -= 1
            self.new_frame_flag = True

    def on_press(self, key):
        if key == keyboard.Key.right:  # Right arrow key
            self.next_frame()
        elif key == keyboard.Key.left:  # Left arrow key
            self.prev_frame()
        elif str(key) == "'q'":  # Press 'q' to exit
            return False

    def play(self):
        listener = keyboard.Listener(on_press=self.on_press)
        listener.start()

        self.show_frame(0)  # Show the first frame

        while True:
            if self.new_frame_flag:
                self.show_frame(self.frame_index)
            if cv2.waitKey(1) & 0xFF == ord('q'):  # Press 'q' to exit
                break

        cv2.destroyAllWindows()

if __name__ == "__main__":
    player = VideoPlayer(rf'F:\data\9.mp4')
    player.play()

python 读取mp4,缓存图片在队列中，这个速度快

就是全部缓存费内存。

import cv2
from collections import deque
from pynput import keyboard


def on_press(key):
    global queue_index, load_new_frame
    if key == keyboard.Key.right:  # Right arrow key
        if queue_index < len(frame_buffer) - 1:
            queue_index += 1
        else:
            load_new_frame = True
    elif key == keyboard.Key.left:  # Left arrow key
        if queue_index > 0:
            queue_index -= 1
        load_new_frame = False
    elif str(key) == "'q'":  # Press 'q' to exit
        return False


listener = keyboard.Listener(on_press=on_press)
listener.start()

if __name__ == '__main__':
    cap = cv2.VideoCapture(r'F:\data\jupiter\err_nos\0912\569\high_res_20230912-151006_SW0031423091207095841748_0_569.mp4')

    deque_len = 10
    frame_buffer = deque(maxlen=deque_len)
    queue_index = -1
    load_new_frame = True
    frame_index = -1
    while cap.isOpened():
        if load_new_frame:
            ret, frame = cap.read()
            if not ret:
                print('video end')
                continue

            frame_index += 1
            cv2.putText(frame, str(frame_index), (20, 70), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2)
            frame_buffer.append(frame)
            queue_index = len(frame_buffer) - 1
            # print('load frame_i', frame_index, "queue_i", queue_index, 'queue len', len(frame_buffer))
            load_new_frame = False

        # Show the current frame
        if len(frame_buffer) > 0:
            view_index = frame_index - len(frame_buffer) + queue_index + 1
            imshow = frame_buffer[queue_index]
            # print('show frame_i', frame_index, "queue_i", queue_index, 'view_i', view_index)
            cv2.imshow('Video', imshow)

        if cv2.waitKey(1) & 0xFF == ord('q'):  # Press 'q' to exit
            break

    # Release the video and close windows
    cap.release()
    cv2.destroyAllWindows()

队列封装成类版本：

import glob
import os
import re

import cv2
from collections import deque

import yaml
from pynput import keyboard

from a2_jupter_check.ip_result.track_tool import get_box_dict, get_color

def show_box(track_id, box, black_image):
    color = get_color(box['sku'])
    cv2.rectangle(black_image, (box['box'][0], box['box'][1]), (box['box'][2], box['box'][3]), color, 1)
    cv2.putText(black_image, f"p:{(box['prob']):.3f}", (box['box'][0], box['box'][1] + 20), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)
    cv2.putText(black_image, "t:" + str(track_id), (box['box'][0], box['box'][1] + 40), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)
    cv2.putText(black_image, "s:" + str(box['sku']), (box['box'][0], box['box'][1] + 60), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)


class VideoPlayer:
    def __init__(self, video_path,video_path2,track_txt):

        self.cap = cv2.VideoCapture(video_path)
        if door_type==1:
            self.cap2 = cv2.VideoCapture(video_path2)
        self.frame_index = -1
        self.total_frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))
        if door_type == 1:
            self.total_frames2 = int(self.cap2.get(cv2.CAP_PROP_FRAME_COUNT))
            self.total_frames=min(self.total_frames,self.total_frames2)

        self.top_dict,  self.bot_dict,  self.end_index = get_box_dict(track_txt)
        self.queue_index = -1
    def save_cls_err(self,cam_id):
        if cam_id==0:
            box_id=0
            for box in self.top_boxes:
                img_crop=self.top_image_o[box['box'][1]:box['box'][3],box['box'][0]:box['box'][2]]
                save_path=f"{cls_dir}/{box['sku']}_{(box['prob']):.2f}_{self.frame_index}_0_{box_id}.jpg"
                cv2.imwrite(save_path,img_crop)
                print('cls top save img',save_path)
                box_id+=1

        elif cam_id==1:
            box_id = 0
            for box in self.bot_boxes:
                img_crop = self.bot_image_o[box['box'][1]:box['box'][3], box['box'][0]:box['box'][2]]
                save_path=f"{cls_dir}/{box['sku']}_{(box['prob']):.2f}_{self.frame_index}_1_{box_id}.jpg"
                cv2.imwrite(save_path, img_crop)
                print('cls bot save img',save_path)
                box_id += 1
    def on_press(self, key):
        global box_offset
        if key == keyboard.Key.right or key == keyboard.Key.space:  # Right arrow key
            if self.queue_index < len(frame_buffer) - 1:
                self.queue_index += 1
            else:
                self.load_new_frame = True
        elif key == keyboard.Key.left:  # Left arrow key
            if self.queue_index > 0:
                self.queue_index -= 1
            self.load_new_frame = False
        elif key == keyboard.KeyCode.from_char('+'):  # Numpad +
            box_offset+=1
            print("box_offset +1", box_offset)
            self.new_frame_flag = True
        elif key == keyboard.KeyCode.from_char('-'):  # Numpad -
            box_offset-=1
            print("box_offset -1", box_offset)
            self.new_frame_flag = True

        elif hasattr(key, 'char') and key.char =='1':
            self.save_cls_err(0)
        elif hasattr(key, 'char') and key.char =='2':
            self.save_cls_err(1)
        elif hasattr(key, 'char') and key.char =='3':
            save_path = f"{det_dir}/{sw_name}_{self.frame_index}_0.jpg"
            cv2.imwrite(save_path, self.top_img_big)
            print('save det err cam 0', save_path)
        elif hasattr(key, 'char') and key.char =='4':
            save_path = f"{det_dir}/{sw_name}_{self.frame_index}_1.jpg"
            cv2.imwrite(save_path, self.bot_img_big)
            print('save det err cam 1',save_path)
        elif str(key) == "'c'":  # 's' key
            print('c')
        elif str(key) == "'d'":  # 's' key
            print('d')
        elif str(key) == "'q'":  # Press 'q' to exit
            return False
        else:
            try:
                if key.char == '/':  # 检查按键字符
                    print('Numpad / key pressed')
                elif key.char=='*':
                    print('Numpad * key pressed')
            except AttributeError:
                pass
    def show_frame(self):
        listener = keyboard.Listener(on_press=self.on_press)
        listener.start()
        self.load_new_frame = True
        while self.cap.isOpened():
            if self.load_new_frame:
                ret, frame = self.cap.read()
                if not ret:
                    print('video end')
                    continue

                self.frame_index += 1
                cv2.putText(frame, str(self.frame_index), (20, 70), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2)
                frame_buffer.append(frame)
                self.queue_index = len(frame_buffer) - 1
                # print('load frame_i', self.frame_index, "queue_i", queue_index, 'queue len', len(frame_buffer))
                self.load_new_frame = False

            # Show the current frame
            if len(frame_buffer) > 0:
                view_index = self.frame_index - len(frame_buffer) + self.queue_index + 1
                imshow = frame_buffer[self.queue_index]
                # print('show frame_i', self.frame_index, "queue_i", queue_index, 'view_i', view_index)
                cv2.imshow('Video', imshow)

            if cv2.waitKey(1) & 0xFF == ord('q'):  # Press 'q' to exit
                break

        # Release the video and close windows
        self.cap.release()
        cv2.destroyAllWindows()
if __name__ == '__main__':

    sbbh='2212311106000040'

    dir_a= r'F:\data\jupiter\err_nos\0912\569'
    dir_a= r'D:\data\err_orders\jupter_pair_09\SW0031423091206302824557'
    dir_a= r'D:\data\err_orders\jupter_pair_09\SW0031423091206323752502'
    cooler_path = rf'F:\data\jupiter\configs/{sbbh}/config/config_cooler.yaml'

    with open(cooler_path, 'r') as file:
        yaml_dict = yaml.safe_load(file)
    img_x=yaml_dict['img_x']
    # img_x = 1920 - 1240 - img_x
    img_y=yaml_dict['img_y']
    top_line=yaml_dict['top_line']
    bot_line=yaml_dict['bot_line']
    img_height, img_width = 744, 1240
    door_type = 1

    cls_dir=dir_a+'/cls'
    det_dir=dir_a+'/det'
    os.makedirs(cls_dir,exist_ok=True)
    os.makedirs(det_dir,exist_ok=True)
    mp4_files = glob.glob(dir_a + '/high_*.mp4')
    # img_x=300+75
    # img_y= 202
    box_offset= 0
    video_path=video_path2=None
    for file in mp4_files:
        if "high_" in file and "_0_" in file:
            video_path=file
        elif "high_" in file and "_1_" in file:
            video_path2=file
        elif "high_" in file and "_2_" in file:
            video_path=file
            door_type = 2
            print("door_type=2")

    if door_type==1 and (video_path is None or video_path2 is None):
        print("cap1 or cap2 is None")
        exit(0)
    txt_path = video_path.replace(".mp4", "_tracks.txt")

    pattern = r'SW\d+'
    match = re.search(pattern, txt_path)
    sw_name = ""
    if match:
        sw_name = match.group()
    if door_type==2:
        img_x = 1920 - 1240 - img_x

    deque_len = 30
    frame_buffer = deque(maxlen=deque_len)


    save_width = 1000
    save_height = save_width *(img_height * 2)  // img_width
    mp4_path=os.path.dirname(txt_path)+'/'+sw_name+'_tracks.mp4'
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # 或者使用 'X264'
    out_writer = cv2.VideoWriter(mp4_path, fourcc, 25.0, (save_width, save_height))
    print('save track mp4 ',mp4_path,'w',save_width,'h',save_height)
    player = VideoPlayer(video_path,video_path2, txt_path)
    player.show_frame()