Kinect关节识别
Posted 81192
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了Kinect关节识别相关的知识,希望对你有一定的参考价值。
基础:添加KinectManager 组件
1)局部关节获取(参考插件场景KinectOverlayDemo1)
要获取局部某一关节及其位置,添加脚本JointOverlayer即可,通过Tracked joint参数可以分别获取到相关关节位置以及坐标。
2)所有关节获取(参考插件场景KinectOverlayDemo2)
获取所有关节,可用脚本SkeletonOverlayer
也可参考如下代码
/// <summary> /// 通过index获取用户的关节,并在每个关节点放置cube /// </summary> /// <param name="index"></param> private void GetBodyJoints(int index) { if(GetDetected()) { long userId = kinectManger.GetUserIdByIndex(index); int jointsCount = kinectManger.GetJointCount(); if(!isJointsCreated) { for (int i = 0; i < jointsCount; i++) { joints.Add(Instantiate(jointGo)); } isJointsCreated = true; } else { for (int i = 0; i < jointsCount; i++) { joints[i].transform.position = kinectManger.GetJointKinectPosition(userId,i); joints[i].name = i.ToString(); } } } }
3)交互功能(参考插件场景KinectOverlayDemo3)
交互功能即为通过手势实现鼠标点击功能需要InteractionManager以及HandOverlayer脚本,通过HandOverlayer/InteractionManager脚本上的参数带texture的参数来实现抓取松开等样式
如果要把手的位置反应到UI上则需要特殊处理,因为我们获取的手的坐标单位为米,反应到unity中只是为数字(比如左手移动0.2m),但是如果分辨率为1920X1080,则手的移动0.2相对于分辨率来说几乎为零,所以需要通过UICamera.ViewportToWorldPoint来转化一下(或者 UICamera.ViewportToScreenPoint)下述脚本为在JointOverlayer中修改的update部分,通过获取手移动的box(范围),然后通过手实际位置转换成相对坐标,然后转化到实际分辨率下的位置。
void Update () { KinectManager manager = KinectManager.Instance; if(manager && manager.IsInitialized() && foregroundCamera) { //backgroundImage.renderer.material.mainTexture = manager.GetUsersClrTex(); if(backgroundImage && (backgroundImage.texture == null)) { backgroundImage.texture = manager.GetUsersClrTex(); } // get the background rectangle (use the portrait background, if available) Rect backgroundRect = foregroundCamera.pixelRect; PortraitBackground portraitBack = PortraitBackground.Instance; if(portraitBack && portraitBack.enabled) { backgroundRect = portraitBack.GetBackgroundRect(); } // overlay the joint int iJointIndex = (int)trackedJoint; if (manager.IsUserDetected()) { for (int i = 0; i < 6; i++) { long userId = manager.GetUserIdByIndex(i); //manager.IsUserTracked(userId); if (manager.IsJointTracked(userId, iJointIndex)) { //MainFunController.theController.OnDebug("Joint tracked\r\n" + GetPara(manager)); overlayObject.gameObject.SetActive(true); Vector3 rightIboxRightBotBack = Vector3.zero, rightIboxRightTopFront = Vector3.zero, rightHandScreenPos = Vector3.zero; bool isrightIboxValid = false; isrightIboxValid = manager.GetRightHandInteractionBox(userId, ref rightIboxRightBotBack, ref rightIboxRightTopFront, isrightIboxValid); if (isrightIboxValid && manager.GetJointTrackingState(userId, (int)KinectInterop.JointType.HandRight) != KinectInterop.TrackingState.NotTracked) { Vector3 rightHandPos = manager.GetJointPosition(userId, (int)KinectInterop.JointType.HandRight); rightHandScreenPos.x = Mathf.Clamp01((rightHandPos.x - rightIboxRightBotBack.x) / (rightIboxRightTopFront.x - rightIboxRightBotBack.x)); rightHandScreenPos.y = Mathf.Clamp01((rightHandPos.y - rightIboxRightBotBack.y) / (rightIboxRightTopFront.y - rightIboxRightBotBack.y)); rightHandScreenPos.z = Mathf.Clamp01((rightIboxRightBotBack.z - rightHandPos.z) / (rightIboxRightBotBack.z - rightIboxRightTopFront.z)); } Vector3 cursorTargetPos = UICamera.ViewportToWorldPoint(rightHandScreenPos); overlayObject.position = Vector2.Lerp(overlayObject.position, cursorTargetPos, 1f); break; } else { overlayObject.gameObject.SetActive(false); //MainFunController.theController.OnDebug("Joint miss"); //isFirstCatch = true; //CollisionEvent.theEvent.OnStop(); } } } else { //debugText.text = "body miss"; } } }
注:应用时建议复制此脚本,并修改名称,然后做自己的修改
以上是关于Kinect关节识别的主要内容,如果未能解决你的问题,请参考以下文章