Enumerate supported pixel formats (VIDIOC_ENUM_FMT). 枚举设备支持的像素格式（VIDIOC_ENUM_FMT）。 Call with `index` starting from 0, incrementing until `EINVAL` is

Enumerate supported frame sizes for a given pixel format (VIDIOC_ENUM_FRAMESIZES). 枚举指定像素格式下支持的帧尺寸（VIDIOC_ENUM_FRAMESIZES）。

Converts a Media Foundation GUID to a RustCV pixel format. This function maps Windows Media Foundation video format GUIDs to the cross-platform `Pixe

Gets the current exposure time. # Returns The exposure time in microseconds, or a default value if unavailable.

Returns a clone of the source reader for control interfaces.

在指定窗口中显示图像

检查是否为空

... 其他 getter ...

Lists all available camera devices connected to the system # Returns A vector of `DeviceInfo` structs containing device name, ID, and backend informa

枚举系统中的摄像头设备

--- 内部辅助函数 (保持不变) ---

Memory-map a V4L2 buffer into user space. 将 V4L2 缓冲区内存映射到用户空间。 Uses `MAP_SHARED` so user space and kernel see the same physical memory (zero-copy). U

Unmap a previously mapped buffer. 解除先前映射的缓冲区。

Negotiates and selects the best video format from available options This function: 1. Gets the presentation descriptor from the MediaSource 2. Extrac

Negotiate the best format based on user config and device capabilities. 根据用户配置和设备能力协商最佳格式。 Strategy: 1. If user specified pixel_format → use it (erro

等待并返回下一帧。 帧数据的生命周期绑定到 `self`（零拷贝借用语义）。

打开设备并初始化流

Opens a camera device for video capture. This method: 1. Initializes Media Foundation if needed 2. Creates a media source for the specified device 3.

Open a V4L2 device node (e.g., "/dev/video0"). 打开 V4L2 设备节点（如 "/dev/video0"）。 Opens with `O_RDWR` (read/write) without `O_NONBLOCK`. Without `O_NONBL

Parses a Media Foundation media type and calculates format score This function extracts video format parameters from an IMFMediaType and calculates a

在图像上绘制文字

Query a single buffer's offset and length (VIDIOC_QUERYBUF). 查询单个缓冲区的偏移量和大小（VIDIOC_QUERYBUF）。 The returned `m.offset` and `length` are needed for `mm

Query device capabilities (VIDIOC_QUERYCAP). 查询设备能力（VIDIOC_QUERYCAP）。 Used to check if the device supports video capture (`V4L2_CAP_VIDEO_CAPTURE`) a

Capture and decode the next frame into a reusable [`Mat`]. 采集并解码下一帧到可复用的 [`Mat`]。 This combines `next_frame()` + decode in a single call. The `Mat` i

简单的最小二乘法实现

在 Mat 上绘制矩形 (In-place) 这是一个手动实现的高性能版本，直接操作 Vec<u8>，避免了任何类型转换。

获取像素数据的切片 (考虑 Stride)

Run a single resolution test with live preview window. 运行单次分辨率测试，带实时预览窗口。 Captures frames for ~5 seconds (or until ESC), returns average FPS. 采集约 5 秒

根据用户配置选择最佳的 AVFoundation 会话预设。

Set a V4L2 control value (VIDIOC_S_CTRL). 设置 V4L2 控制参数值（VIDIOC_S_CTRL）。 Common uses: - Disable `exposure_dynamic_framerate` (id=0x009a0903, value=0)

Set the video format (VIDIOC_S_FMT). 设置视频格式（VIDIOC_S_FMT）。 The driver may adjust the requested values to the nearest supported ones. The returned str

Set the frame rate via VIDIOC_S_PARM. 通过 VIDIOC_S_PARM 设置帧率。 Frame rate is expressed as a fraction: `timeperframe = numerator / denominator`. For 30f

启动帧采集。

Starts the video capture stream. This initiates the first sample request, which triggers the callback pipeline.

Start the capture session. 启动采集会话。

Row step in bytes. 行步长（字节）。

停止帧采集（不销毁管线，可重新 start）。

Stops the video capture stream. Flushes any pending samples and prevents further callbacks.

Stop the capture session. 停止采集会话。

Stop video streaming. 停止视频流。

Stop video streaming (VIDIOC_STREAMOFF). 停止视频流（VIDIOC_STREAMOFF）。 All buffers are automatically dequeued by the driver after this call. 此调用后驱动会自动将所有缓

Start video streaming (VIDIOC_STREAMON). 启动视频流（VIDIOC_STREAMON）。 All buffers must be queued before calling this. 调用前所有缓冲区必须已入队。

Copy the raw data into an [`OwnedFrame`] that can outlive the camera borrow. 将原始数据拷贝到 [`OwnedFrame`]，使其可以超出摄像头的借用生命周期。 Use this when you need to: - S

将 rustcv-core 的 FourCC 转换为 v4l 的 FourCC 用于请求设备设置格式

Waits for a sample to become available and returns it. Uses a semaphore to track available frames. The semaphore's permits accumulate when frames arr

等待按键

辅助：YUYV -> BGR (保留之前的实现)

专门为 image crate 优化的 YUYV -> RGB8 (R,G,B, R,G,B...)

Called when a new sample is available from the source reader.

基于遥测数据简单的健康评估

Benchmark 4: Detailed frame statistics (custom, not criterion). 基准测试 4：详细帧统计（自定义，非 criterion）。 Runs after criterion benchmarks to print a human-reada

Benchmark 1: Raw DQBUF throughput (zero-copy, no decode). 基准测试 1：原始 DQBUF 吞吐量（零拷贝，无解码）。 Measures the maximum frame rate achievable by the capture eng

Benchmark 3: Multi-resolution comparison. 基准测试 3：多分辨率对比。 Compares capture FPS at 480p, 720p, and 1080p. 对比 480p、720p、1080p 下的采集 FPS。

Benchmark 2: DQBUF + MJPEG decode (turbojpeg). 基准测试 2：DQBUF + MJPEG 解码（turbojpeg）。 Measures the full VideoCapture::read() pipeline including JPEG dec

Set the number of mmap buffers to allocate. 设置要分配的 mmap 缓冲区数量。 Recommended values by target fps: - 30 fps: 5 buffers (default) - 60 fps: 8 buffers -

设置缓冲区数量 (默认 3)

Calculates buffer size based on pixel format. Returns a tuple of (stride, total_size) where: - stride: bytes per row - total_size: total buffer size

Number of channels (always 3 for BGR). 通道数（BGR 固定为 3）。

Image width (number of columns). 图像宽度（列数）。

Access the raw pixel data (zero-copy). 访问原始像素数据（零拷贝）。 The format of this data depends on [`pixel_format()`](Self::pixel_format): - `Mjpeg`: JPEG-comp

Decode this frame into a BGR [`Mat`]. 将此帧解码为 BGR [`Mat`]。 This is a convenience method that allocates a new Mat. For zero-allocation decoding, use [`

辅助：获取首选后端类型

Returns the default MSMF driver as a trait object. This is a convenience function for applications that need a dynamically dispatched driver instance

为了方便直接使用，提供一个默认实例

Dequeue a frame from the driver (blocking). 从驱动出队一帧（阻塞）。 This is the **hot path** — called once per frame. No heap allocations, no poll(), single ioc

创建一个空的 Mat (通常用于作为输出 buffer)