关键词: 物体识别, 图像处理, 快速响应矩阵码, 类海尔特征

Abstract: There are many objects with different shapes and colors in home environment. It is difficult to separate the target from the complicated background image using the traditional image processing method. QR Code tag was designed for each object, so the identification process of object was changed into the identification process of QR Code. This method can reduce the calculation work, and improve the identification precision and speed. Manipulation information (such as grasp force, grasp pose, etc.) can be simultaneously recorded in the QR Code. The robot can know how to manipulate the target by decoding the QR Code. The QR Code detection and localization were realized by the trained cascade of boosted classifiers based on Haarlike features. Experiments show that the proposed method can recognize the target with high adaptability and robustness.

Key words: objects identification, image processing, quick response code, haar-like feature