基于MLKit的Android人脸识别应用开发实践

艾克华

于 2025-04-16 17:02:45 发布

阅读量393

点赞数 5

文章标签： android 人脸识别 kotlin

本文链接：https://blog.csdn.net/weixin_37391237/article/details/147281604

版权

基于MLKit的Android人脸识别应用开发实践

https://gitee.com/wenhua512/face-recognition

1. 项目概述

1.1 功能特点

实时人脸检测与跟踪
人脸特征提取与识别
自动/手动采集模式
人脸数据管理
相机参数优化

1.2 技术选型

MLKit人脸检测
MediaPipe人脸网格
CameraX相机框架
Room数据库
Kotlin协程

2. 核心实现

2.1 人脸检测配置

private fun createFaceDetector(): FaceDetector {
    val options = FaceDetectorOptions.Builder()
        .setPerformanceMode(FaceDetectorOptions.PERFORMANCE_MODE_FAST)
        .setContourMode(FaceDetectorOptions.CONTOUR_MODE_ALL)
        .setLandmarkMode(FaceDetectorOptions.LANDMARK_MODE_ALL)
        .setClassificationMode(FaceDetectorOptions.CLASSIFICATION_MODE_ALL)
        .setMinFaceSize(0.15f)
        .build()
    return FaceDetection.getClient(options)
}

2.2 相机预览实现

private fun startCamera() {
    val cameraProviderFuture = ProcessCameraProvider.getInstance(this)
    
    cameraProviderFuture.addListener({
        val cameraProvider = cameraProviderFuture.get()
        val preview = Preview.Builder()
            .setTargetResolution(targetResolution)
            .build()
            .also {
                it.setSurfaceProvider(binding.viewFinder.surfaceProvider)
            }
            
        val imageAnalyzer = ImageAnalysis.Builder()
            .setTargetResolution(targetResolution)
            .setBackpressureStrategy(ImageAnalysis.STRATEGY_BLOCK_PRODUCER)
            .build()
            .also {
                it.setAnalyzer(cameraExecutor, FaceAnalyzer())
            }
            
        val cameraSelector = CameraSelector.DEFAULT_FRONT_CAMERA
        
        try {
            cameraProvider.unbindAll()
            cameraProvider.bindToLifecycle(
                this, cameraSelector, preview, imageAnalyzer
            )
        } catch(exc: Exception) {
            Log.e(TAG, "相机绑定失败", exc)
        }
    }, ContextCompat.getMainExecutor(this))
}

2.3 人脸特征提取

private suspend fun extractFaceFeatures(bitmap: Bitmap): FloatArray? {
    return withContext(Dispatchers.IO) {
        try {
            val faceBitmap = cropFace(bitmap, faceMesh.boundingBox)
            val processedBitmap = ImagePreprocessor.process(
                faceBitmap,
                shouldNormalizeLighting = true,
                shouldEnhanceContrast = true,
                shouldEqualizeHistogram = true
            )
            faceFeatureExtractor.extractFeatures(processedBitmap)
        } catch (e: Exception) {
            Log.e(TAG, "特征提取失败", e)
            null
        }
    }
}

3. 性能优化

3.1 图像处理优化

使用YUV格式处理图像
实现图像缓存机制
优化人脸裁剪算法

3.2 内存管理

private fun cleanupResources() {
    try {
        if (::cameraExecutor.isInitialized && !cameraExecutor.isShutdown) {
            cameraExecutor.shutdown()
        }
        
        currentFaceBitmap?.recycle()
        currentFaceBitmap = null
        latestBitmap?.recycle()
        latestBitmap = null
        
        collectedFeatures.clear()
    } catch (e: Exception) {
        Log.e(TAG, "资源清理失败", e)
    }
}

3.3 多线程处理

private val processingMutex = Mutex()
private val processIntervalMs = 1000L

private suspend fun processFrame(bitmap: Bitmap) {
    if (!processingMutex.tryLock()) return
    
    try {
        val currentTime = System.currentTimeMillis()
        if (currentTime - lastProcessTime < processIntervalMs) return
        
        // 处理逻辑
        lastProcessTime = currentTime
    } finally {
        processingMutex.unlock()
    }
}

4. 用户体验优化

4.1 采集模式切换

private fun updateCollectionModeUI() {
    binding.apply {
        autoCollectLayout.visibility = if (isAutoCollectMode) View.VISIBLE else View.GONE
        manualCollectLayout.visibility = if (isAutoCollectMode) View.GONE else View.VISIBLE
        
        if (isAutoCollectMode) {
            tvStatus.text = "自动采集模式：请保持人脸在框内..."
        } else {
            Toast.makeText(this@FaceAutoCollectActivity, 
                "手动采集模式：请保持人脸在框内...", 
                Toast.LENGTH_SHORT).show()
        }
    }
}

4.2 人脸质量检测

private fun isValidFaceMesh(faceMesh: FaceMesh, imageProxy: ImageProxy): Boolean {
    // 检查人脸角度
    val faceAngle = calculateFaceAngle(faceMesh)
    if (abs(faceAngle) > MAX_FACE_ANGLE) {
        showHint("请保持正面朝向摄像头")
        return false
    }
    
    // 检查人脸大小
    val faceSize = calculateFaceSize(faceMesh, imageProxy)
    if (faceSize < MIN_FACE_SIZE || faceSize > MAX_FACE_SIZE) {
        showHint("请调整与摄像头的距离")
        return false
    }
    
    return true
}