基于支持向量机和神经网络的实时警报系统

Real Time Alarm System Based on Support Vector Machines and Neural Net

根据芯片生产线等场所的需要和现有人工监控手段以及国外基于支持向量机相关产品的缺陷,该文利用图像的二维矩不变量理论,将实时图像转换成灰度图像后,用CANNY算子作边缘检测,并计算边缘检测图像的二维不变矩,再利用支持向量机的支持向量回归理论对二维不变矩进行训练和识别,提出了一种基于支持向量机与神经网络的实时警报系统的设计算法,给出了算法实例和结果。从实验仿真结果和实际运行情况来看,算法的效果是令人满意的。

According to the needs of microprocessor product line etc. and disfigurement of manpower control measure in existence and foreign correlative devices based on SVM, this article brings forward an real time alarm system design algorithm and implementation based on SVM and NN. The algorithm uses two dimension constant moment theory, converting the real time image to the grayscale intensity image, finding edges in the grayscale intensity image using CANNY arithmetic operators, and calculating two dimension constant moment of the edges image, making use of SVR of SVM training and identify two dimension constant moment. It gives examples of the algorithm and experimental result which is both reasonable and satisfactory.