MacOS下基于AVFoundation的Python图像捕获函数静默失败求助

修复后的完整代码 #

import os
import AVFoundation
import Foundation
import time
from threading import Semaphore


def capture_image(unique_id, dest_filename):
    # 初始化信号量，用于等待异步捕获完成
    semaphore = Semaphore(0)
    # 设置AVFoundation捕获会话
    session = AVFoundation.AVCaptureSession.alloc().init()
    # 获取所有设备并过滤视频设备（替代废弃的devicesWithMediaType_）
    devices = AVFoundation.AVCaptureDevice.devices()
    device = None
    for dev in devices:
        if dev.hasMediaType_(AVFoundation.AVMediaTypeVideo) and dev.uniqueID() == unique_id:
            device = dev
            break
    if device is None:
        raise ValueError(f"No camera found with unique ID: {unique_id}")
    
    # 创建设备输入并处理错误
    input_session, error = AVFoundation.AVCaptureDeviceInput.deviceInputWithDevice_error_(device, None)
    if error is not None:
        raise RuntimeError(f"Failed to create capture input: {error}")
    
    output_session = AVFoundation.AVCaptureStillImageOutput.alloc().init()
    # 应用JPG编码输出设置
    output_settings = {AVFoundation.AVVideoCodecKey: AVFoundation.AVVideoCodecTypeJPEG}
    output_session.setOutputSettings_(output_settings)
    
    session.addInput_(input_session)
    session.addOutput_(output_session)
    session.startRunning()

    # 等待相机启动就绪（可根据实际情况调整时长）
    time.sleep(1)

    # 定义捕获完成回调
    def capture_completion(buffer, error):
        if error is not None:
            print(f"Capture error: {error}")
            semaphore.release()
            return
        # 从缓冲区提取JPG数据并写入文件
        image_data = AVFoundation.AVCaptureStillImageOutput.jpegStillImageNSDataRepresentation_(buffer)
        success, write_error = image_data.writeToFile_options_error_(dest_filename, Foundation.NSDataWritingAtomic, None)
        if not success and write_error is not None:
            print(f"Failed to write image: {write_error}")
        semaphore.release()

    # 异步捕获图像
    connection = output_session.connectionWithMediaType_(AVFoundation.AVMediaTypeVideo)
    output_session.captureStillImageAsynchronouslyFromConnection_completionHandler_(connection, capture_completion)
    
    # 等待捕获逻辑执行完毕
    semaphore.acquire()
    
    # 释放会话资源
    session.stopRunning()
    session.release()
    output_session.release()
    input_session.release()
    
    print(f'writing {dest_filename}')
    return dest_filename


if __name__ == '__main__':
    # 确保目标目录存在
    os.makedirs('/tmp/images', exist_ok=True)
    images_dict = {
        '0x143141300bda5829': '/tmp/images/image_1.jpg',
        '0x143141100bda5829': '/tmp/images/image_2.jpg',
        '0x143200000bda5829': '/tmp/images/image_3.jpg',
        '0x143300000bda5829': '/tmp/images/image_4.jpg',
        '0x143121200bda5829': '/tmp/images/image_5.jpg',
    }
    list_of_image_files_written = []
    for the_unique_id, dest_path in images_dict.items():
        try:
            captured_path = capture_image(the_unique_id, dest_path)
            if os.path.exists(captured_path):
                list_of_image_files_written.append(os.path.basename(captured_path))
        except Exception as e:
            print(f"Failed to capture image for {the_unique_id}: {e}")
    print(f'{list_of_image_files_written=}')

关键问题修复说明 #

1. 缺失Foundation框架导入 #

原代码使用NSData和NSDataWritingAtomic但未导入Foundation框架，直接导致运行时失败。添加import Foundation即可解决该问题。

2. 异步捕获未等待完成 #

captureStillImageAsynchronouslyFromConnection_completionHandler_是异步执行的，原代码调用后立即停止会话，此时捕获逻辑还未执行就被终止，导致图像无法写入。使用Semaphore信号量等待回调执行完毕后再停止会话，确保捕获流程完成。

3. 输出设置未生效 #

原代码定义了output_settings但未调用output_session.setOutputSettings_()绑定到输出会话，导致JPG编码设置未生效。

4. 废弃API替换 #

AVCaptureDevice.devicesWithMediaType_在新版本AVFoundation中已废弃，改用devices()获取所有设备后，通过hasMediaType_过滤视频设备。

5. 错误处理补充 #

添加了设备输入创建失败、捕获错误、文件写入错误的处理逻辑，便于排查具体失败原因。

6. PyCharm未解析引用问题 #

PyCharm对PyObjC的动态Objective-C绑定支持有限，出现的未解析引用提示不影响运行。若要消除提示，可安装类型存根包：

pip install pyobjc-stubs

7. 摄像头权限检查 #

确保运行程序的进程（如PyCharm）拥有摄像头访问权限：

• 打开系统设置 → 隐私与安全性 → 摄像头
• 确认PyCharm已在列表中并勾选权限

额外注意事项 #

• 多次调用捕获函数时，需确保前一个会话资源完全释放后再创建新会话，避免设备资源冲突。
• time.sleep()的时长可根据相机启动速度调整，无需固定2秒。
• 提前创建目标目录，避免因目录不存在导致写入失败。

内容的提问来源于stack exchange，提问作者mcgregor94086