一种基于FPGA的实时彩色图像目标分类方法

An FPGA-based Real-time Color Image Classification Method

提出了一种基于FPGA的实时彩色图像目标分类算法.为了在不同光照条件下标定目标颜色,算法采用了一种统计椭球模型;该算法在YUV空间只用18位而非24位数据建立了3D颜色查找表.这样解决了以往算法中存在的目标颜色体重叠和占用存储空间大的问题,并且提高了算法的分类准确性.同时,为了进一步减轻机器人主CPU的运算负担,利用FPGA技术对该算法进行了硬件实现,极大地提高了视觉系统的性能.该方法在智能移动机器人和工业计算机视觉系统研究领域具有广泛的应用前景.

Based on the technology of FPGA, a real-time object classification method in color image is presented. In order to calibrate the object color in different lighting conditions, a kind of statistic ellipsoidal model is used. In this method, a 3-D Color Look-up Table （CLUT） is built in which only 18 bits are applied to represent one kind of color, instead of conventional 24 bits. So this method resolves the problem of object overlapping in color space, and has the merits of higher classification accuracy and lower memory cost than traditional ones. Moreover, this method is implemented on FPGA, which can highly reduce the CPU computation burden and remarkably improve the performance of the vision system. This method is validated by the applications on mobile robot and industrial vision system.