如何实现类似FaceID的iPhone用户注意力检测？求Vision Framework相关方案

核心实现方案（基于Vision Framework） #

Vision Framework本身没有直接提供注视检测的API，但可以通过人脸关键点检测，分析瞳孔与眼睛轮廓的相对位置来实现类似功能。以下是完整的Swift代码示例：

import UIKit
import AVFoundation
import Vision

class GazeDetectionVC: UIViewController, AVCaptureVideoDataOutputSampleBufferDelegate {
    private var captureSession: AVCaptureSession!
    private let detectionQueue = DispatchQueue(label: "com.yourapp.gazedetection")
    
    override func viewDidLoad() {
        super.viewDidLoad()
        setupCameraSession()
    }
    
    private func setupCameraSession() {
        captureSession = AVCaptureSession()
        captureSession.sessionPreset = .medium
        
        // 获取前置摄像头
        guard let frontCam = AVCaptureDevice.default(.builtInWideAngleCamera, for: .video, position: .front) else {
            fatalError("前置摄像头不可用")
        }
        
        // 创建摄像头输入
        guard let camInput = try? AVCaptureDeviceInput(device: frontCam) else {
            fatalError("无法初始化摄像头输入")
        }
        
        if captureSession.canAddInput(camInput) {
            captureSession.addInput(camInput)
        }
        
        // 设置视频输出与代理
        let videoOutput = AVCaptureVideoDataOutput()
        videoOutput.setSampleBufferDelegate(self, queue: detectionQueue)
        if captureSession.canAddOutput(videoOutput) {
            captureSession.addOutput(videoOutput)
        }
        
        captureSession.startRunning()
    }
    
    // 处理摄像头帧数据
    func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
        guard let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else { return }
        
        // 创建人脸关键点检测请求
        let landmarkRequest = VNDetectFaceLandmarksRequest(completionHandler: handleLandmarkResults)
        let imageHandler = VNImageRequestHandler(cvPixelBuffer: pixelBuffer, orientation: .leftMirrored)
        
        do {
            try imageHandler.perform([landmarkRequest])
        } catch {
            print("检测失败: \(error.localizedDescription)")
        }
    }
    
    // 处理人脸关键点结果
    private func handleLandmarkResults(request: VNRequest, error: Error?) {
        guard let faceObservations = request.results as? [VNFaceObservation], let face = faceObservations.first else {
            print("未检测到人脸")
            return
        }
        
        guard let faceLandmarks = face.landmarks, let leftEye = faceLandmarks.leftEye, let rightEye = faceLandmarks.rightEye else {
            print("未检测到眼睛关键点")
            return
        }
        
        // 计算双眼中心点
        let leftEyeCenter = calculateEyeCenter(eye: leftEye)
        let rightEyeCenter = calculateEyeCenter(eye: rightEye)
        
        // 判断是否注视屏幕
        let isGazingAtScreen = checkGazeValidity(leftEyeCenter: leftEyeCenter, leftEyeBounds: leftEye.boundingBox,
                                                 rightEyeCenter: rightEyeCenter, rightEyeBounds: rightEye.boundingBox)
        
        DispatchQueue.main.async {
            // 这里可以更新UI或执行业务逻辑
            print(isGazingAtScreen ? "正在注视屏幕" : "未注视屏幕")
        }
    }
    
    // 计算眼睛中心点（基于关键点平均）
    private func calculateEyeCenter(eye: VNFaceLandmarkRegion2D) -> CGPoint {
        let points = eye.normalizedPoints
        let avgX = points.reduce(0) { $0 + $1.x } / CGFloat(points.count)
        let avgY = points.reduce(0) { $0 + $1.y } / CGFloat(points.count)
        return CGPoint(x: avgX, y: avgY)
    }
    
    // 判断注视有效性（通过瞳孔与眼睛边界的相对位置）
    private func checkGazeValidity(leftEyeCenter: CGPoint, leftEyeBounds: CGRect,
                                   rightEyeCenter: CGPoint, rightEyeBounds: CGRect) -> Bool {
        // 阈值可根据实际测试调整，控制检测灵敏度
        let horizontalThreshold: CGFloat = 0.25
        
        // 左眼水平范围判断
        let leftEyeCenterX = leftEyeBounds.midX
        let leftValidRange = (leftEyeCenterX - leftEyeBounds.width * horizontalThreshold)...(leftEyeCenterX + leftEyeBounds.width * horizontalThreshold)
        
        // 右眼水平范围判断
        let rightEyeCenterX = rightEyeBounds.midX
        let rightValidRange = (rightEyeCenterX - rightEyeBounds.width * horizontalThreshold)...(rightEyeCenterX + rightEyeBounds.width * horizontalThreshold)
        
        return leftValidRange.contains(leftEyeCenter.x) && rightValidRange.contains(rightEyeCenter.x)
    }
}

关键注意事项 #

• 权限配置：必须在Info.plist中添加NSCameraUsageDescription字段，说明相机使用用途，否则无法获取摄像头权限。
• 设备兼容性：搭载TrueDepth摄像头的设备（iPhone X及以后机型）能提供更精准的眼睛关键点数据；普通前置摄像头的检测精度有限。
• 阈值调整：checkGazeValidity中的horizontalThreshold需要根据实际测试调整，适配不同用户的面部特征。

官方文档参考 #

苹果官方文档中可重点关注以下内容：

• VNDetectFaceLandmarksRequest类：用于发起人脸关键点检测请求
• VNFaceObservation的landmarks属性：包含人脸各部位（眼睛、眉毛、嘴巴等）的关键点集合
• VNFaceLandmarkRegion2D类：提供关键点的归一化坐标与边界信息

内容的提问来源于stack exchange，提问作者XCozi