iOS平台实时摄像头运动检测应用开发实现方法咨询

嘿，我来帮你搞定iOS上的实时摄像头运动检测需求！咱们从核心思路到具体代码一步步拆解，确保你能快速落地：

一、核心实现思路 #

要完成实时摄像头的运动检测，本质上是捕获摄像头实时帧→分析帧间变化/识别目标移动→触发运动事件。根据你的需求（检测人、动物的运动），可以分两种方案：

• 轻量通用方案：帧差法（快速检测任何运动，适合低性能设备）
• 精准目标方案：Core ML目标检测+跟踪（只关注人/动物的移动，精度更高）

二、具体实现步骤 #

1. 先搞定摄像头权限 #

iOS对隐私权限要求严格，第一步必须申请摄像头权限：

• 在Info.plist中添加NSCameraUsageDescription，填写权限说明（比如“需要访问摄像头进行运动检测”）
• 代码中主动请求权限：

import AVFoundation

func requestCameraPermission() {
    AVCaptureDevice.requestAccess(for: .video) { granted in
        DispatchQueue.main.async {
            if granted {
                // 权限通过，启动摄像头会话
                self.setupCaptureSession()
            } else {
                // 提示用户开启权限
                print("摄像头权限被拒绝，请在设置中开启")
            }
        }
    }
}

2. 搭建摄像头实时捕获会话 #

用AVCaptureSession来获取摄像头的实时帧数据：

var captureSession: AVCaptureSession!
var lastGrayscaleFrame: CVPixelBuffer?

func setupCaptureSession() {
    captureSession = AVCaptureSession()
    captureSession.sessionPreset = .medium // 平衡分辨率和性能
    
    // 获取后置摄像头
    guard let camera = AVCaptureDevice.default(.builtInWideAngleCamera, for: .video, position: .back) else {
        print("无法访问后置摄像头")
        return
    }
    
    // 添加输入
    guard let input = try? AVCaptureDeviceInput(device: camera) else { return }
    if captureSession.canAddInput(input) {
        captureSession.addInput(input)
    }
    
    // 添加输出（获取实时帧）
    let videoOutput = AVCaptureVideoDataOutput()
    videoOutput.setSampleBufferDelegate(self, queue: DispatchQueue(label: "cameraQueue"))
    if captureSession.canAddOutput(videoOutput) {
        captureSession.addOutput(videoOutput)
    }
    
    // 启动会话
    captureSession.startRunning()
}

3. 实现帧处理逻辑 #

遵循AVCaptureVideoDataOutputSampleBufferDelegate协议，处理每一帧：

extension YourViewController: AVCaptureVideoDataOutputSampleBufferDelegate {
    func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
        guard let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else { return }
        
        // 方案1：帧差法检测通用运动
        detectMotionWithFrameDifference(currentPixelBuffer: pixelBuffer)
        
        // 方案2：Core ML检测人/动物的移动
        // detectTargetMotionWithCoreML(currentPixelBuffer: pixelBuffer)
    }
}

三、两种检测方案详解 #

方案1：帧差法（轻量快速） #

通过对比连续两帧的灰度图像素差异，判断是否有运动：

func detectMotionWithFrameDifference(currentPixelBuffer: CVPixelBuffer) {
    // 转灰度图减少计算量
    let currentGrayscale = convertToGrayscale(pixelBuffer: currentPixelBuffer)
    
    guard let lastFrame = lastGrayscaleFrame else {
        lastGrayscaleFrame = currentGrayscale
        return
    }
    
    // 计算帧间像素差异占比
    let differenceRatio = calculatePixelDifference(buffer1: lastFrame, buffer2: currentGrayscale)
    
    // 设定阈值（比如5%，可根据需求调整）
    if differenceRatio > 0.05 {
        DispatchQueue.main.async {
            // 触发运动检测事件，比如更新UI或发送通知
            print("检测到运动！")
        }
    }
    
    lastGrayscaleFrame = currentGrayscale
}

// 辅助：转灰度图
func convertToGrayscale(pixelBuffer: CVPixelBuffer) -> CVPixelBuffer {
    let ciImage = CIImage(cvPixelBuffer: pixelBuffer)
    let grayscaleFilter = CIFilter(name: "CIPhotoEffectNoir")!
    grayscaleFilter.setValue(ciImage, forKey: kCIInputImageKey)
    guard let outputImage = grayscaleFilter.outputImage else { return pixelBuffer }
    
    // 转换回CVPixelBuffer
    let context = CIContext()
    let outputBuffer = pixelBuffer.copy() as! CVPixelBuffer
    context.render(outputImage, to: outputBuffer)
    return outputBuffer
}

// 辅助：计算像素差异占比
func calculatePixelDifference(buffer1: CVPixelBuffer, buffer2: CVPixelBuffer) -> Double {
    let width = CVPixelBufferGetWidth(buffer1)
    let height = CVPixelBufferGetHeight(buffer1)
    var differentPixels = 0
    let totalPixels = width * height
    
    CVPixelBufferLockBaseAddress(buffer1, .readOnly)
    CVPixelBufferLockBaseAddress(buffer2, .readOnly)
    
    let baseAddr1 = CVPixelBufferGetBaseAddress(buffer1)!
    let baseAddr2 = CVPixelBufferGetBaseAddress(buffer2)!
    
    for y in 0..<height {
        for x in 0..<width {
            let pixel1 = baseAddr1.load(fromByteOffset: y*CVPixelBufferGetBytesPerRow(buffer1) + x, as: UInt8.self)
            let pixel2 = baseAddr2.load(fromByteOffset: y*CVPixelBufferGetBytesPerRow(buffer2) + x, as: UInt8.self)
            if abs(Int(pixel1) - Int(pixel2)) > 30 { // 像素差异阈值
                differentPixels += 1
            }
        }
    }
    
    CVPixelBufferUnlockBaseAddress(buffer1, .readOnly)
    CVPixelBufferUnlockBaseAddress(buffer2, .readOnly)
    
    return Double(differentPixels) / Double(totalPixels)
}

方案2：Core ML目标检测+跟踪（精准识别目标） #

如果只需要检测人、动物的运动，可以用预训练的Core ML模型（比如YOLOv8、MobileNetV2）先识别目标，再跟踪目标的位置变化：

1. 下载预训练模型（比如从Apple的Core ML Models库选目标检测模型），导入Xcode
2. 实现目标检测和跟踪：

import CoreML
import Vision

// 提前加载模型
lazy var detectionRequest: VNCoreMLRequest = {
    do {
        let model = try VNCoreMLModel(for: YOLOv8n().model) // 替换成你的模型
        let request = VNCoreMLRequest(model: model) { [weak self] request, error in
            self?.processDetectionResults(request: request)
        }
        request.imageCropAndScaleOption = .scaleFill
        return request
    } catch {
        fatalError("加载模型失败：\(error)")
    }
}()

func detectTargetMotionWithCoreML(currentPixelBuffer: CVPixelBuffer) {
    let handler = VNImageRequestHandler(cvPixelBuffer: currentPixelBuffer, orientation: .up)
    do {
        try handler.perform([detectionRequest])
    } catch {
        print("检测失败：\(error)")
    }
}

var lastTargetPositions: [String: CGPoint] = [:]

func processDetectionResults(request: VNRequest) {
    guard let results = request.results as? [VNRecognizedObjectObservation] else { return }
    
    // 过滤出人和动物（根据模型的标签，比如"person"、"dog"、"cat"等）
    let targetResults = results.filter { result in
        let topLabel = result.labels.first?.identifier ?? ""
        return ["person", "dog", "cat", "bird"].contains(topLabel)
    }
    
    DispatchQueue.main.async {
        for result in targetResults {
            let targetLabel = result.labels.first?.identifier ?? "unknown"
            let currentCenter = CGPoint(x: result.boundingBox.midX, y: result.boundingBox.midY)
            
            // 对比上一帧的位置，判断是否移动
            if let lastPosition = self.lastTargetPositions[targetLabel] {
                let distance = sqrt(pow(currentCenter.x - lastPosition.x, 2) + pow(currentCenter.y - lastPosition.y, 2))
                if distance > 0.02 { // 移动距离阈值，可调整
                    print("检测到\(targetLabel)移动！")
                }
            }
            
            self.lastTargetPositions[targetLabel] = currentCenter
        }
    }
}

四、优化建议 #

• 性能优化：降低摄像头分辨率（比如用.low preset）、减少检测帧率（比如每2帧检测一次）、用Metal加速像素计算
• 功耗控制：不需要检测时暂停摄像头会话，避免持续耗电
• 后台限制：iOS后台下摄像头只能短暂工作，若需后台检测，需申请background modes中的Audio, AirPlay, and Picture in Picture（因为摄像头和音频权限绑定）

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