特征提取在电力图像识别系统中的应用

Application of Feature Extraction in Power Image Recognition System

电力图像识别系统存在图像信息资源丢失的问题,导致图像识别精度低,系统运行时间长。文中设计了一种基于特征提取的电力图像识别系统。在硬件部分,FPGA板和摄像头连接板集成在一块电路板上,外部时钟通过专用全局时钟引脚MCLK进入FPGA,图像数据和同步信号通过摄像头链路接口协议传输。软件部分提取电源图像的颜色、形状和纹理特征,并根据空间关系形成特征网络。测试结果表明,基于特征提取的电力图像识别系统可以提高图像识别的准确性,缩短网络训练的时间。

The power image recognition system has the problem of loss of image information resources,resulting in low image recognition accuracy and long system running time.A power image recognition system based on feature extraction is designed.In the hardware part,the FPGA board and the camera connection board are integrated on one circuit board.The external clock enters the FPGA through the dedicated global clock pin MCLK,and the image data and synchronization signals are transmitted through the camera link interface protocol.The software part extracts the color,shape and texture features of the power image,and forms a feature network according to the spatial relationship.A power image recognition model based on feature extraction is designed.As a sample,each predicted frame is labeled according to its position and category relative to the real frame,thereby obtaining the label value to predict the target position and category.The test results show that the power image recognition system based on feature extraction can improve the accuracy of image recognition and shorten the network training time.