机器人视觉中物体识别实验系统设计.docVIP

机器人视觉中物体识别实验系统设计 摘 要 作为机电一体化产品的典型代表，机器人在逐渐走向智能化。而通过视觉来对周围环境中的物体进行识别的技术是机器人智能化的重要标志。具有“视觉”且能够识别物体的机器人可以获取外部世界图像，并从中识别出感兴趣的物体，根据需要作出反应。这对机器人的工作和未来机器人的发展具有重要的意义。且物体识别不仅仅可以用于机器人，其他各行各业如交通，国防等也都用到该项技术。而物体识别中最具代表的是人脸的检测，本次设计为从图像或视频中检测出人脸。 本次设计的方法为首先搜集大量正负样本，并对样本进行归一化操作。然后用opencv图像库根据矩形特征，Adaboost算法训练出级联分类器，编写主程序读取图片或者从摄像头获取的视频的每一帧，对图像（帧）进行扫描，检测出人脸的个数（如果存在）、位置和大小，并标记出来予以显示。 最后需要对设计成品进行检验，将含有人脸的图片或摄像头获取的内容提供给程序，观察程序的输出结果（对人脸的检测标记情况）是否与预期一致，确保它能够满足要求，并由此发现设计的不足之处，以便在以后的工作中予以改进。 关键字 物体识别，人脸检测，OpenCV，训练，分类器 OBJECT RECOGNITION OF ROBOT VISION EXPERIMENTAL SYSTEM DESIGN ＡＢＳＴＲＡＣＴ As a typical of mechannical and electronics production, robots have been going toward intelligence gradually. The technology of recognizing objects in the environment is an important symbol of robots intelligence.The robots that have vision and recognize objects are able to get the picture of external environment,and recognize the objects that they are interested in.The they response according to need.this is a very important meaning for the work of robot and the development of future robot.And this technology can not only be used in robot,but also other fields such as traffic and national defense.Face detection is a typical of objects recognition.So the content of this design is to detect faces from the image or video. The method is as follows.First,I collect lots of positive and negtive samples ,and normalize them.Then,I train the cascade classifier by opencv according to Haar feather and Adaboost Algorithm,and programme to read images or every frame of video from the camera.Then the programme will scan the iamge (frame),detect the number of faces(if any),the position and size,and it will mark it and show it. At last,we should test the work.we can provide images or videos got from the camera to the programme,then we look at and analyse what the programme show out(the result that it detect and mark faces).And the result should be compared with the situation we expect.Having done all of the