一种新的计算机视觉算法

受蚂蚁金服计算机视觉项目组委托,为了提高计算机视觉的识别效率和准确率,利用机器学习的思想开发自动进化图像学习机,将MATHEMATICA Cloud作为改学习机的计算云、搜索云,将学习函数和决策函数嵌入到学习机中,从而实现由学习机录入图像并导入MATHEMATICA Cloud进行计算和搜索的功能,利用搜索结果来锻炼学习机的学习函数,利用计算结果来锻炼学习机的决策函数,从而实现高效率的识别速度与高准确率的识别效果.基于蚂蚁金服提供的验算实例进行仿真实验,并从计算速度、收敛情况、识别精度等方面同现在常用的识别算法进行对比,结果显示:新算法的图像识别能力颇佳,与传统算法相比,新算法具有更快的计算速度、绝对收敛性和远高于传统算法的识别精度.

A New Computer Vision Algorithm

By the commission of the computer vision project of ant gold service,in order to improve the recognition efficiency of the computer vision and accuracy,using the idea of machine learning to develop automatic evolution image learning machine,this article uses Mathematica cloud as the computing cloud,search cloud of this modified learning machine,makes learning function and decision function embed into the learning machine in order to achieve the calculation and search by inputting image and importing Mathematica Cloud to learning machine.By using the search results to exercise the learning function of the learning machine,by using the calculated results to exercise decision-making function of learning machine,in order to achieve high efficiency recognition speed and high accuracy rate of recognition effect.Based on the calculating example provided by ant gold service,the simulation experiment has been carried on and compared with the common recognition algorithm from calculation speed,convergence and identification precision.The results show that image recognition ability is quite good in the new algorithm proposed in this paper.Compared with the traditional algorithm,this algorithm has faster computing speed,absolute convergence,and much higher than traditional algorithm in recognition accuracy.