人们与Kinect对抗 [英] People counter with Kinect

查看:57
本文介绍了人们与Kinect对抗的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!

问题描述


我正在使用一个kinect设备制作人员计数器原型,顶部安装在零售商店的天花板上。我在考虑使用以下算法来计算人数:


- 从传感器获取深度图;


- 根据深度信息,查找靠近传感器的头顶;


- 将头部顶部提取为斑点,找到它们的质心并跟踪它们;


- 每次blob越过图像中的某个区域时递增/递减计数器。


我知道kinect提供跟踪人的功能,我可以将kinect挂在墙上使用了内置的跟踪功能,但我们的原型是用于在繁忙的交通区域(在购物中心内),我不知道kinect内的
跟踪机制是否足够可靠,可以计算大一群人,如果它有能力处理遮挡。 


我也知道有人会建议使用普通的摄像机,但他们对光和图像条件敏感,当kinect深度传感器不是。


请有人可以告诉我,我的方法是否可行?有没有人曾尝试做过类似的事情?


祝你好运,


Tony Belloni

解决方案

Tony。


我认为你的方法比KinectSDK跟踪机制更好地计算交通繁忙地区的人数。首先,用户索引(使用索引深度图)是3位大小,因此可以通过Kinect同时跟踪不超过7个人。而且,KinectSDK
不仅可以追踪人物。它跟踪一把椅子,一条裙子,一台电脑(我们真的很惊讶而不是看到它)......也许这意味着任何物体都有精神,但它不符合科学的观点。


所以,我认为,你可以更好地使用没有内置kinect功能的深度图。


我希望你觉得这很有用。


< 。p>大沙&NBSP;


Hi,

I am working on a prototype of a people counter using one kinect device, top-mounted on the ceiling of a retail store. I am thinking in use the following algorithm to count people:

- Get the depth map from the sensor;

- Based on depth information, find the top of heads which are closer to the sensor;

- Extract the top of heads as blobs, find their Center of Mass and track them;

- Increment/decrement a counter every time a blob cross a certain area in the image.

I know that kinect provide functions to track people and that I could mount the kinect on the wall and used the built-in tracking features, but our prototype is intendend to be used in a heavy traffic area (inside a shoping mall) and I don't know if the tracking mechanism inside kinect would be reliable enough to count large groups of people and if it is capable dealing with occlusion. 

I also know that someone would suggest to use regular video cameras, but they are sensitive to light and image conditions, when the kinect depth sensor is not.

Please, could someone tell me if my approach is feasible ? Has anyone tried to do something like that before ?

Best regards,

Tony Belloni

解决方案

Hi, Tony.

I think your approach is better than KinectSDK tracking mechanism to count people in a heavy traffic area. Firstly, user index ( using indexep depth map) is 3 bit size, so no more than 7 people can be tracked simultaneously via Kinect. Moreover, KinectSDK can track not only people. It track a chair, a skirt, a computer ( we was realy surprised than saw it)... Maybe it means that any object has a spirit, but it doesn't fit to scientific point of view.

So, I think, you can better use depth map without built-in kinect features.

I hope you find this helpful.

Dasha. 


这篇关于人们与Kinect对抗的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!

查看全文
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆