基于opencv对高空拍摄视频消抖处理

import cv2import timeimport numpy as npimport os  class Stable:    # 处理视频文件路径    __video_path = None     # surf 特征提取    __surf = {        # surf算法        'surf': None,        # 提取的特征点        'kp': None,        # 描述符        'des': None,        # 过滤后的特征模板        'template_kp': None    }     # capture    __capture = {        # 捕捉器        'cap': None,        # 视频大小        'size': None,        # 视频总帧        'frame_count': None,        # 视频帧率        'fps': None,        'video': None    }     # 配置    __config = {        # 要保留的最佳特征的数量        'key_point_count': 5000,        # Flann特征匹配        'index_params': dict(algorithm=0, trees=5),        'search_params': dict(checks=50),        'ratio': 0.5,        'frame_count': 9999    }     # 当前处理帧数    __current_frame = 0     # 需要处理帧数    __handle_count = 0     # 处理时间    __handle_timer = {        'init': 0,        'handle': 0,        'read': 0,        'key': 0,        'matrix': 0,        'flann': 0,        'perspective': 0,        'write': 0,        'other': 0,    }     # 帧队列    __frame_queue = None     # 需要写入的帧队列    __write_frame_queue = None     # 特征提取列表    __surf_list = []     def __init__(self):        pass     # 初始化capture    def __init_capture(self):        self.__capture['cap'] = cv2.VideoCapture(self.__video_path)        self.__capture['size'] = (int(self.__capture['cap'].get(cv2.CAP_PROP_FRAME_WIDTH)),                                  int(self.__capture['cap'].get(cv2.CAP_PROP_FRAME_HEIGHT)))         self.__capture['fps'] = self.__capture['cap'].get(cv2.CAP_PROP_FPS)         self.__capture['video'] = cv2.VideoWriter(self.__video_path.replace('.', '_stable.'),                                                  cv2.VideoWriter_fourcc(*"mp4v"),                                                  self.__capture['fps'],                                                  self.__capture['size'])         self.__capture['frame_count'] = int(self.__capture['cap'].get(cv2.CAP_PROP_FRAME_COUNT))         self.__handle_count = min(self.__config['frame_count'], self.__capture['frame_count'])     # 初始化surf    def __init_surf(self):         st = time.time()        self.__capture['cap'].set(cv2.CAP_PROP_POS_FRAMES, 0)        state, first_frame = self.__capture['cap'].read()          self.__capture['cap'].set(cv2.CAP_PROP_POS_FRAMES, self.__capture['frame_count'] - 20)        state, last_frame = self.__capture['cap'].read()         self.__surf['surf'] = cv2.xfeatures2d.SURF_create(self.__config['key_point_count'], 1, 1, 1, 1)         # nfeatures：默认为0，要保留的最佳特征的数量。特征按其分数排名（在SIFT算法中按局部对比度排序）        # nOctaveLayers：默认为3，金字塔每组(Octave)有多少层。3是D. Lowe纸中使用的值。        # contrastThreshold：默认为0.04，对比度阈值，用于滤除半均匀（低对比度）区域中的弱特征。阈值越大，检测器产生的特征越少。        # edgeThreshold：默认为10，用来过滤边缘特征的阈值。注意，它的意思与contrastThreshold不同，edgeThreshold越大，滤出的特征越少（保留更多特征）。        # sigma：默认为1.6，高斯金字塔中的σ。如果使用带有软镜头的弱相机拍摄图像，则可能需要减少数量。         self.__surf['kp'], self.__surf['des'] = self.__surf['surf'].detectAndCompute(first_frame, None)        kp, des = self.__surf['surf'].detectAndCompute(last_frame, None)         # 快速临近匹配        flann = cv2.FlannBasedMatcher(self.__config['index_params'], self.__config['search_params'])        matches = flann.knnMatch(self.__surf['des'], des, k=2)         good_match = []        for m, n in matches:            if m.distance < self.__config['ratio'] * n.distance:                good_match.append(m)         self.__surf['template_kp'] = []        for f in good_match:            self.__surf['template_kp'].append(self.__surf['kp'][f.queryIdx])         self.__capture['cap'].set(cv2.CAP_PROP_POS_FRAMES, 0)         self.__handle_timer['init'] = int((time.time() - st) * 1000)         print("[INFO] init time:{}ms".format(self.__handle_timer['init']))     # 初始化 队列    def __init_data(self):        pass     # 初始化    def __init(self):        self.__init_capture()        self.__init_surf()        self.__init_data()     # 处理    def __process(self):         self.__current_frame = 1         while True:             if self.__current_frame > self.__handle_count:                break             start_time = time.time()             # 抽帧            success, frame = self.__capture['cap'].read()            self.__handle_timer['read'] = int((time.time() - start_time) * 1000)             if not success: return             # 计算            frame = self.detect_compute(frame)             # 写帧            st = time.time()            self.__capture['video'].write(frame)            self.__handle_timer['write'] = int((time.time() - st) * 1000)             self.__handle_timer['handle'] = int((time.time() - start_time) * 1000)             self.__current_frame += 1             self.print_handle_time()     # 视频稳像    def stable(self, path):        self.__video_path = path        self.__init()        self.__process()     # 打印耗时    def print_handle_time(self):        print(            "[INFO] handle frame:{}/{} time:{}ms(read:{}ms key:{}ms flann:{}ms matrix:{}ms perspective:{}ms write:{}ms)".                format(self.__current_frame,                       self.__handle_count,                       self.__handle_timer['handle'],                       self.__handle_timer['read'],                       self.__handle_timer['key'],                       self.__handle_timer['flann'],                       self.__handle_timer['matrix'],                       self.__handle_timer['perspective'],                       self.__handle_timer['write']))     # 特征点提取    def detect_compute(self, frame):         frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)         # 计算特征点        st = time.time()        kp, des = self.__surf['surf'].detectAndCompute(frame_gray, None)        self.__handle_timer['key'] = int((time.time() - st) * 1000)         # 快速临近匹配        st = time.time()        flann = cv2.FlannBasedMatcher(self.__config['index_params'], self.__config['search_params'])        matches = flann.knnMatch(self.__surf['des'], des, k=2)        self.__handle_timer['flann'] = int((time.time() - st) * 1000)         # 计算单应性矩阵        st = time.time()        good_match = []        for m, n in matches:            if m.distance < self.__config['ratio'] * n.distance:                good_match.append(m)         p1, p2 = [], []        for f in good_match:            # 存在与模板特征点中            if self.__surf['kp'][f.queryIdx] in self.__surf['template_kp']:                p1.append(self.__surf['kp'][f.queryIdx].pt)                p2.append(kp[f.trainIdx].pt)         H, _ = cv2.findHomography(np.float32(p2), np.float32(p1), cv2.RHO)        self.__handle_timer['matrix'] = int((time.time() - st) * 1000)         # 透视变换        st = time.time()        output_frame = cv2.warpPerspective(frame, H, self.__capture['size'], borderMode=cv2.BORDER_REPLICATE)        self.__handle_timer['perspective'] = int((time.time() - st) * 1000)         return output_frame  s = Stable() s.stable('video/test10.mov')