TANG Minan, LUO Yinhang, WANG Chenyu, ZHANG Kaiyue
（School of Automation and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China）
Abstract： An improved visual background extractor (ViBe) algorithm combined with perceptual hash algorithm, frame difference and frequency-tuned (FT) saliency detection algorithm is proposed to solve the problem that the traditional ViBe algorithm is susceptible to ghost, background noise and other factors in the foreground detection. Firstly, the saliency image and the frame difference image of current frame are used to extract moving target area in real time to eliminate ghost phenomenon. Secondly, sample set information is exploited to analyze the complexity of the background which adjusts the radius threshold of pixel segmentation adaptively. Finally, the number of connected domain pixels in the target image is used for the second foreground detection, and fragmented pixels and noise are filtered out to obtain the real target image. The experimental data show that the F-Measure value is maintained at a high level while detected images have a high accuracy rate. The experimental results show that the improved algorithm can relatively fast remove the ghost phenomenon and improve the adaptability in dynamic scenes.
Key words： image processing； ViBe algorithm； perceptual hash algorithm； frame difference method； saliency detection algorithm
References
［1］ CHEN H Y, QIE L Z, YANG D D, et al. Visual background extraction algorithm based on superpixel information feedback. Acta Optica Sinica, 2017, 37（7）： 715001.
［2］ MAHALINGAM T, SUBRAMONIAM M. CBFD: a refined W4+ cluster-based frame difference approach for efficient moving object detection. Soft Computing, 2019, 23（21）： 10661-10679.
［3］ SENGAR S S, MUKHOPADHYAY S. Moving object detection based on frame difference and W4. Signal, Image and Video Processing, 2017, 11（7）： 1357-1364.
［4］ SANCHES S R R, SEMENTILLE A C, AGUILAR I A, et al. Recommendations for evaluating the performance of background subtraction algorithm for surveillance systems. Multimedia Tools and Applications, 2021, 80（3）： 4421-4454.
［5］ RAJIV K, ARUN K, GAJENDRA S. Optical flow-based weighted magnitude and direction histograms for the detection of abnormal visual events using combined classifier. International Journal of Cognitive Informatics and Natural Intelligence, 2021, 15（3）： 12-30.
［6］ GUO S N, WU C X, LIN X, et al. A discrimination method of ghost and static target based on improved ViBe. Data Science and Industrial Internet, 2018, 1（1）： 1-12.
［7］ YANG L, LU J Z. Optimization and implementation of pedestrian detection system in surveillance video based on DSP. Transducer and Microsystem Technologies, 2021, 40 （1）： 91-93.
［8］ LIU Y D, ZENG T W, CHEN D B, et al. Adaptive moving target detection algorithm based on regional similarity information. Computer Engineering, 2020, 46（3）： 273-279．
［9］ LIANG J C, WANG X P, WANG Q S. An improved ViBe algorithm for illumination mutations. Journal of Measurement Science and Instrumentation, 2021, 12(2): 184-191.
［10］ WU S B, WANG H F, YAN B Y, et al. VOCs Leakage automatic identification algorithm based on ViBe model. Control Engineering of China, 2020, 27（11）： 1970-1974.
［11］ SINGH R P, SHARMA P, MADARKAR J. Compute-extensive background subtraction for efficient ghost suppression. IEEE Access, 2019, 7: 130180-130196.
［12］ YE G, LI Y B, MA Z X, et al. End-to-end aluminum strip surface defects detection and recognition method based on ViBe. Journal of Zhejiang University （Engineering Science), 2020, 54（10）： 1906-1914.
［13］ WANG W, WANG X P, LIANG J C. An improved ViBe algorithm based on adaptive detection of moving targets.Journal of Measurement Science and Instrumentation, 2020, 11（2）： 126-134.
［14］ HE L F, LIU Y L, ZHONG Y, et al. Video foreground target extraction algorithm in complex background. Laser & Optoelectronics Progress, 2020, 57（16）： 161013.
［15］ LIU Z H, HUANG J, LIU W Q, et al. Prospect detection based on improved visual background extraction algorithm. Journal of Chinese Computer Systems, 2020, 41（4）： 758-762.
［16］ MIN W D, GUO X G, HAN Q. An improved ViBe algorithm and its application in traffic video processing. Optics and Precision Engineering, 2017, 25（3）： 806-811.
［17］ TONG Y, SUO J D, REN S L. A ship detection algorithm of remote sensing technology based on ship saliency in candidate regions. Electronics Optics & Control, 2021, 28（2）： 48-52.
［18］ XU Y P, JI H B, ZHANG W B. Ghost detection and removal based on two-layer background model and histogram similarity. Sensors, 2020, 20（16）： 1-23.
［19］ HUO Y L, QIN M, SONG Y L. Moving target detection method based on edge character and multiple frame difference. Infrared Technology, 2017, 39（2）： 157-162.
［20］ LIU J J, LIN H K. Adaptive threshold moving target detection algorithm based on ViBe. Laser & Optoelectronics Progress, 2022, 59(2): 144-153.
基于ViBe算法的鬼影消除与自适应前景检测策略
汤旻安， 罗引航， 王晨雨， 张凯悦
（兰州交通大学 自动化与电气工程学院， 甘肃 兰州 730070）
摘要：为了解决传统视频背景提取算法（Visual background extractor， ViBe）前景检测时易受“鬼影”、 背景噪声等因素干扰的问题， 结合感知哈希算法、 帧差法与频域协调显著性检测算法， 提出了一种改进的ViBe算法。 首先， 利用当前帧的显著性图像与帧差图像， 实时确定运动目标区域， 消除“鬼影”现象。 其次， 根据样本集信息得到背景复杂度， 自适应调节像素分割的半径阈值。 最后， 通过目标图像中连通域像素数量进行前景二次检测， 滤除背景像素与噪声， 得到真实目标图像。 实验数据显示， 检测后的图像在保持有较高准确率的同时F-measure值也维持在较高水平， 证明改进算法可以快速消除鬼影现象， 提高动态场景下的适应性。
关键词：图像处理； 视频背景提取算法； 感知哈希算法； 帧差法； 显著性检测算法
引用格式：TANG Minan, LUO Yinhang, WANG Chenyu, et al． Ghost elimination and adaptive foreground detection strategy based on ViBe algorithm． Journal of Measurement Science and Instrumentation， 2023， 14（3）： 299-305. DOI： 10．3969／j．issn．1674-8042．2023．03．006O
[full text view]
