图像目标抗遮挡视觉差补偿算法仿真  

Simulation of Anti Occlusion Visual Difference Compensation Algorithm for Image Objects

作　　者：赵海燕[1] 杜丽娟[1] 刘琨[1] 肖琳[1] ZHAO Hai-yan;DU Li-juan;LIU Kun;XIAO Lin(College of Applied Science&Technology,Beijing Union University,Beijing 100012,China)

机构地区：[1]北京联合大学应用科技学院,北京100012

出　　处：《计算机仿真》2025年第1期244-248,共5页Computer Simulation

基　　金：2021年度北京市教委科研计划项目(KM202111417002)。

摘　　要：由于场景复杂、光线变化等因素的影响,图像中的目标物体可能被遮挡,导致视差的失真和误差,从而影响对图像目标的精准跟踪。为了有效解决上述问题,提出图像目标抗遮挡视觉差补偿算法。将图像转移到HSI颜色空间中,检测并消除图像中存在的噪声;将去噪后的图像输入YOLOv3网络中,检测并定位目标区域;根据人眼视觉特性建立二维视觉差异直方图,结合LDR算法对目标区域展开增强处理,通过灰度补偿方法对目标区域展开灰度补偿,避免过度增强现象的发生,实现目标抗遮挡视觉差补偿。实验结果表明,所提算法的图像去噪效果好、目标定位精度高、补偿效果好、补偿效率高。Due to the complexity of the scenes,changes in lighting,and other factors,the target objects in the image may be obscured,resulting in distortion and errors in parallax,which in turn affects the accurate tracking of image targets.To effectively solve this problem,this article presented an anti-occlusion visual difference compensation algorithm for image object.At first,we transformed the image into HSI color space.After that,we detected and eliminated noise from the image.Moreover,we inputted the denoised image into YOLOv3 network to detect and locate the target area.According to human visual characteristics,we established a two-dimensional visual difference histogram and combined with LDR algorithm to enhance the target area.Through grayscale compensation,we compensated the target area,thus avoiding excessive enhancement.Finally,we achieved the anti-occlusion visual difference compensation.The experimental results show that the proposed algorithm has good denoising effect,high target positioning accuracy,good compensation effect,and high compensation efficiency.

关 键 词：颜色空间 目标定位 二维视觉差异直方图 视觉差补偿 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]