融入帧间差分法的深度学习目标识别仿真研究

Research on simulation of deep learning target recognition based on inter-frame difference method

目标检测与识别是数字图像处理实验的创新性实验项目。传统的目标识别算法可以识别目标类型,但不能识别目标的位置信息,且对相同目标的识别率较低。该文设计了一种基于帧间差分法的深度学习目标识别算法,即在深度学习理论构架下,将帧间差分法融入识别过程,补充增强候选框分割图像,通过NMS算法对候选框进行筛选。仿真结果表明,该算法在识别目标种类的同时还能对目标在图像中的位置进行精确标定,并可判断目标是否处于运动状态,具有较高的识别率。

Target detection and recognition are an innovative experimental project in digital image processing experiment.Traditional target recognition algorithm can recognize the type of target,but it can’t recognize the location information of the target,and the recognition rate of the same target is low.A deep learning target recognition algorithm based on frame difference method is designed.In the framework of deep learning theory,the frame difference method is applied to the recognition process,the enhanced candidate frame is segmented and the candidate frame is filtered through the NMS algorithm.The simulation results show that the algorithm can’t only recognize the type of the target,but also accurately calibrate the position of the target in the image.It can also judge whether the target is in a moving state and has a high recognition rate.