这段时间再做安防相关的硬件设备定制，涉及到了小区的业主和流动人员人员登记、闸机管理和小区内部摄像头等等。
这里面用到了人脸检测、识别，人脸检测用了虹软的算法，虹软的文档还是比较全的这里不多做介绍。这里主要说下android自带的人脸检测和人脸框的绘制。
先来看看人脸识别框特效：
这个例子使用Camera API实现的并不是Camera2 API，原因是发现2019年了还有的设备居然不支持Camera2的人脸识别，为了通用就用了Camera API。
以下为主要代码
上代码：
1、下来是处理人脸检测回调方法
//进行相机初始化
cameraHelper = CameraControllerHelper.Builder()
.previewViewSize(Point(camera_surfaceview.measuredWidth, camera_surfaceview.measuredHeight))
.rotation(mBaseAc.windowManager.defaultDisplay.rotation)
.specificCameraId(cameraId)
.isMirror(isMirror())
.previewOn(camera_surfaceview)
.cameraListener(this@CameraFaceFm)
//设置人脸检测回调方法
.setFaceDetectionListener(object : Camera.FaceDetectionListener {
override fun onFaceDetection(p0: Array<out Camera.Face>?, p1: Camera?) {
if (faceRectView != null) {
faceRectView.clearFaceInfo()
faceRectView.addFaceInfo(p0!!.toList())
}
}
})
.build()
val p = getScreenPoint()
DrawFaceHelper.cameraWidth = p.x.toFloat()
DrawFaceHelper.cameraHeight = p.y.toFloat()
DrawFaceHelper.isBackCameraId = isBackCameraId()
cameraHelper.init()
cameraHelper.start()
2、相机的启动、参数设置和人脸检测
if (mCamera != null) {
return
}
//若指定了相机ID且该相机存在，则打开指定的相机
if (specificCameraId != null) {
mCameraId = specificCameraId!!
} else {
//相机数量为2则打开1,1则打开0,相机ID 1为前置，0为后置
mCameraId = Camera.getNumberOfCameras() - 1
}
//没有相机
if (mCameraId == -1) {
cameraListener?.onCameraError(Exception("camera not found"))
return
}
if (mCamera == null) {
mCamera = Camera.open(mCameraId)
}
displayOrientation = getCameraOri(rotation)
mCamera!!.setDisplayOrientation(displayOrientation)
try {
val parameters = mCamera!!.parameters
parameters.previewFormat = ImageFormat.NV21
//预览大小设置
previewSize = parameters.previewSize
val supportedPreviewSizes = parameters.supportedPreviewSizes
if (supportedPreviewSizes != null && supportedPreviewSizes.size > 0) {
previewSize = currentPreviewSize
}
parameters.setPreviewSize(previewSize!!.width, previewSize!!.height)
pictureSize = parameters.pictureSize
val supportedPicviewSizes = parameters.supportedPreviewSizes
if (supportedPicviewSizes != null && supportedPicviewSizes.size > 0) {
pictureSize = currentPictureSize
}
parameters.setPictureSize(pictureSize!!.width, pictureSize!!.height)
//对焦模式设置
val supportedFocusModes = parameters.supportedFocusModes
if (supportedFocusModes != null && supportedFocusModes.size > 0) {
if (supportedFocusModes.contains(Camera.Parameters.FOCUS_MODE_CONTINUOUS_PICTURE)) {
parameters.focusMode = Camera.Parameters.FOCUS_MODE_CONTINUOUS_PICTURE
} else if (supportedFocusModes.contains(Camera.Parameters.FOCUS_MODE_CONTINUOUS_VIDEO)) {
parameters.focusMode = Camera.Parameters.FOCUS_MODE_CONTINUOUS_VIDEO
} else if (supportedFocusModes.contains(Camera.Parameters.FOCUS_MODE_AUTO)) {
parameters.focusMode = Camera.Parameters.FOCUS_MODE_AUTO
}
}
setCameraParameters(parameters)
mCamera!!.setPreviewTexture(previewDisplayView!!.getSurfaceTexture())
mCamera!!.setPreviewCallback(this)
mCamera!!.startPreview()
//注意这段，这里是设置人脸检测监听和启动
if (mFaceDetectionListener != null) {
mCamera!!.setFaceDetectionListener(mFaceDetectionListener)
mCamera!!.startFaceDetection()
}
if (autoFocusCallback != null) {
mCamera!!.autoFocus(autoFocusCallback)
}
cameraListener?.onCameraOpened(mCamera!!, mCameraId, displayOrientation, isMirror)
} catch (e: Exception) {
if (cameraListener != null) {
cameraListener.onCameraError(e)
} else {
e.printStackTrace()
}
}
}
启动相机这块就不多做介绍，以为网上现成的文档还是很多的。
##下来才是重头戏！！！
3、在FaceRectView的onDraw方法内进行人脸框显示
override fun onDraw(canvas: Canvas) {
super.onDraw(canvas)
if (faceRectList != null && faceRectList.isNotEmpty()) {
DrawFaceHelper.drawFaceRect(canvas, faceRectList[0].rect, color, 3)
//不断进行刷新，实现人脸框动画更新
postInvalidate()
} else {
DrawFaceHelper.resetTimerInfo()
}
}
这里需要说明下，相机返回的Face信息里面的Rect的坐标并不是view的真实坐标，这个需要转换。转换方式见 android人脸检测点位置转换
4、用DrawFaceHelper中的drawFaceRect绘制人脸框
/**
* 绘制数据信息到view上，若 [DrawInfo.getName] 不为null则绘制 [DrawInfo.getName]
*
* @param canvas 需要被绘制的view的canvas
* @param drawInfo 绘制信息
* @param color rect的颜色
* @param faceRectThickness 人脸框线条粗细
*/
fun drawFaceRect(canvas: Canvas?, drawInfo: Rect?, color: Int, faceRectThickness: Int) {
var displayOrientation: Int = if (isBackCameraId) { 90 } else { 270 }
//将返回的人脸信息，转为view可识别的信息
val matrix = CameraUtils.prepareMatrix(isBackCameraId, displayOrientation, cameraWidth.toInt(), cameraHeight.toInt())
val rectF = RectF(drawInfo)
matrix.mapRect(rectF)
if (canvas == null || drawInfo == null) {
return
}
if (timer == null) {
initTimerInfo()
}
val paint = Paint()
paint.style = Paint.Style.STROKE //设置空心
paint.strokeWidth = faceRectThickness.toFloat()
val spic = 100
val rect = Rect(rectF.left.toInt() - spic, rectF.top.toInt() - spic, rectF.right.toInt() + spic, rectF.bottom.toInt() + spic)
val width = rect.right - rect.left
val height = rect.bottom - rect.top
//比较小的设置为半径
val radius = if (height > width) { width / 2 } else { height / 2 }
val cx = (rect.left + rect.right) / 2
val cy = (rect.top + rect.bottom) / 2
//绘制遮罩层
val maskPaint = Paint()
//整个相机预览界面进行遮罩，透明度用来设置遮罩半透明
canvas.saveLayerAlpha(0f, 0f, cameraWidth, cameraHeight, maskAlpha, Canvas.ALL_SAVE_FLAG)
maskPaint.color = Color.BLACK
//整个相机预览界面进行遮罩
canvas.drawRect(Rect(0, 0, cameraWidth.toInt(), cameraHeight.toInt()), maskPaint)
//重叠区域进行处理
//只在源图像和目标图像不相交的地方绘制【源图像】，相交的地方根据目标图像的对应地方的alpha进行过滤，
//目标图像完全不透明则完全过滤，完全透明则不过滤
maskPaint.xfermode = PorterDuffXfermode(PorterDuff.Mode.SRC_OUT)
maskPaint.color = Color.BLACK
canvas.drawCircle(cx.toFloat(), cy.toFloat(), radius.toFloat(), maskPaint)
//自定义人脸框样式
paint.flags = Paint.ANTI_ALIAS_FLAG //抗锯齿
paint.color = color
//设置正方形，以便下面设置弧形为圆的弧形
rect.left = cx - radius
rect.top = cy - radius
rect.right = cx + radius
rect.bottom = cy + radius
val rectF0 = getRectF(faceOffset0, rect)
paint.strokeWidth = 4f
paint.alpha = faceAlpha0
canvas.drawArc(rectF0, getStartAngle(faceStartAngle0, false), faceSweepAngle0, useCenter, paint)
val rectF1 = getRectF(faceOffset1, rect)
paint.strokeWidth = 8f
paint.alpha = faceAlpha1
canvas.drawArc(rectF1, getStartAngle(faceStartAngle1, true), faceSweepAngle1, useCenter, paint)
val rectF2 = getRectF(faceOffset2, rect)
paint.strokeWidth = 3f
paint.alpha = faceAlpha2
canvas.drawArc(rectF2, getStartAngle(faceStartAngle2, false), faceSweepAngle2, useCenter, paint)
paint.alpha = faceAlpha3
val rectF3 = getRectF(faceOffset3, rect)
paint.strokeWidth = 9f
canvas.drawArc(rectF3, getStartAngle(faceStartAngle3, true), faceSweepAngle3, useCenter, paint)
val rectF4 = getRectF(faceOffset3, rect)
paint.strokeWidth = 9f
canvas.drawArc(rectF4, getStartAngle(faceStartAngle4, true), faceSweepAngle4, useCenter, paint)
}
/**
* 设置圆环偏移量
* @param startAngle 角度
* @param clockwise 是否顺时针
*/
private fun getStartAngle(startAngle:Float, clockwise: Boolean):Float {
val angle = if (clockwise) { startAngle + timeIndex } else { startAngle - timeIndex}
return angle % 360
}
private fun getRectF(offset: Int, rect: Rect): RectF {
val rectF = RectF(rect)
rectF.left = rectF.left - offset
rectF.top = rectF.top - offset
rectF.right = rectF.right + offset
rectF.bottom = rectF.bottom + offset
return rectF
}
这就是实现的大致流程。
代码里面只支持了显示一张人脸的检测方式。如有需要显示多张人脸的话请各位童鞋开动脑力自己解决！
完整代码请戳这里：FaceCamera
扫一扫
4390
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
