Design And Implementation Of Blob Recognition Software Based On Optical Multi-touch System

With the development of computer vision, multi-touch technology has become one of the most popular issues in the domain of man-computer interaction. As a new type of input unit, optics-based multi-touch system has some features such as multi-blob recognition and big screen. It enjoys vast potential and has been widely applied in hotel, exhibition, publicity of government, ect.The application of multi-touch system has also encouraged the development and improvement of blob recognition software. Whether software can be utilized to improve the adaptability of system and the experience of users is an important factor to multi-touch system. At present, most blob recognition software which support optical multi-touch system have poor performances on the adaptability of changing background and the tracking of fast-moving objects. And most of software are integrated and with low portability.To solve these problems, this paper represented the construction of Frustrated Total Internal Reflection (FITR)-based and Laser Light Plane (LLP)-based multi-touch system on the angle of hardware with the "Multi-touch Platform of Marine Information Technology Laboratory in Ocean University of China". The key process and technology of blob recognition were then detailed. After that, this paper analyzed how to use the dual camera to implement blob recognition in a optical multi-touch system. This paper also showed an application of depth information in blob recognition software.The precise recognition of blob is the core technology of optical control system. To solve the problem that changing background will disturb blob extraction, which exists in many optical multi-touch system, this paper presented an average background model-based blob-detecting algorithm after analyzing some well-known algorithm. This algorithm updated the background gradually using the averaged pixels, and recognized blob using a self-adapting threshold set by improved Nobuyuki Otsu (OTSU) algorithm. The algorithm can eliminate the effect of background varying with time; therefore it can be used to improve the accuracy and adaptality of the system.The main innovations of this paper are:1) An efficient and self-adapting object-extracting algorithm and a method to obtain the segmentation threshold. This function can improve the adaptability of system, the efficiency of blob-tracking and the experiences of users.2) Obtaining the depth information of a moving target with dual camera. These extend the scope of the blob recognition software and enrich the applications of blob recognition software and even optical multi-touch system.