A method for moving object recognition and tracking in the intelligent traffic monitoring system is presented. For the shortcomings and deficiencies of the frame-subtraction method, a redundant discrete wavelet transf...A method for moving object recognition and tracking in the intelligent traffic monitoring system is presented. For the shortcomings and deficiencies of the frame-subtraction method, a redundant discrete wavelet transform （RDWT） based moving object recognition algorithm is put forward, which directly detects moving objects in the redundant discrete wavelet transform domain. An improved adaptive mean-shift algorithm is used to track the moving object in the follow up frames. Experimental results show that the algorithm can effectively extract the moving object, even though the object is similar to the background, and the results are better than the traditional frame-subtraction method. The object tracking is accurate without the impact of changes in the size of the object. Therefore the algorithm has a certain practical value and prospect.展开更多
Compressive sensing is a revolutionary idea proposed recently to achieve much lower sampling rate for signals.In the image application with limited resources the camera data can be stored and processed in compressed f...Compressive sensing is a revolutionary idea proposed recently to achieve much lower sampling rate for signals.In the image application with limited resources the camera data can be stored and processed in compressed form.An algorithm for moving object and region detection in video using a compressive sampling is developed.The algorithm estimates motion information of the moving object and regions in the video from the compressive measurements of the current image and background scene.The algorithm does not perform inverse compressive operation to obtain the actual pixels of the current image nor the estimated background.This leads to a computationally efficient method and a system compared with the existing motion estimation methods.The experimental results show that the sampling rate can reduce to 25% without sacrificing performance.展开更多
Moving object detection is one of the challenging problems in video monitoring systems, especially when the illumination changes and shadow exists. Amethod for real-time moving object detection is described. Anew back...Moving object detection is one of the challenging problems in video monitoring systems, especially when the illumination changes and shadow exists. Amethod for real-time moving object detection is described. Anew background model is proposed to handle the illumination varition problem. With optical flow technology and background subtraction, a moving object is extracted quickly and accurately. An effective shadow elimination algorithm based on color features is used to refine the moving obj ects. Experimental results demonstrate that the proposed method can update the background exactly and quickly along with the varition of illumination, and the shadow can be eliminated effectively. The proposed algorithm is a real-time one which the foundation for further object recognition and understanding of video mum＇toting systems.展开更多
A novel cast shadow detection approach was proposed.A stereo vision system was used to capture images instead of traditional single camera.It was based on an assumption that cast shadows were on a special plane.The im...A novel cast shadow detection approach was proposed.A stereo vision system was used to capture images instead of traditional single camera.It was based on an assumption that cast shadows were on a special plane.The image obtained from one camera was inversely projected to the plane and then transformed to the view from another camera.The points on the plane shared the same position between original image and the transformed image.As a result,the cast shadows can be detected.In order to improve the efficiency of cast shadow detection and decrease computational complexity,the obvious object areas in CIELAB color space were removed and the potential shadow areas were obtained.Experimental results demonstrate that the proposed approach can detect cast shadows accurately even under various illuminations.展开更多
A novel moving object detection method was proposed in order to adapt the difficulties caused by intermittent object motion,thermal and dynamic background sequences.Two groups of complementary Gaussian mixture models ...A novel moving object detection method was proposed in order to adapt the difficulties caused by intermittent object motion,thermal and dynamic background sequences.Two groups of complementary Gaussian mixture models were used.The ghost and real static object could be classified by comparing the similarity of the edge images further.In each group,the multi resolution Gaussian mixture models were used and dual thresholds were applied in every resolution in order to get a complete object mask without much noise.The computational color model was also used to depress illustration variations and light shadows.The proposed method was verified by the public test sequences provided by the IEEE Change Detection Workshop and compared with three state-of-the-art methods.Experimental results demonstrate that the proposed method is better than others for all of the evaluation parameters in intermittent object motion sequences.Four and two in the seven evaluation parameters are better than the others in thermal and dynamic background sequences,respectively.The proposed method shows a relatively good performance,especially for the intermittent object motion sequences.展开更多
行人作为非刚性物体,对它的视觉特征进行有效的不变表示是提高识别效果的关键。在自然视觉场景中,运动行人通常会发生尺度、背景、姿态等变化,这对用现有技术提取这些不规则特征造成阻碍。针对该问题,基于哺乳动物视网膜神经结构特性,...行人作为非刚性物体,对它的视觉特征进行有效的不变表示是提高识别效果的关键。在自然视觉场景中,运动行人通常会发生尺度、背景、姿态等变化,这对用现有技术提取这些不规则特征造成阻碍。针对该问题,基于哺乳动物视网膜神经结构特性,探究运动行人不变性识别问题,并提出一种适用于视觉场景的运动行人检测神经网络(MPDNN)。MPDNN包括2个神经模块:突触前网络和突触后网络。其中,突触前网络感知表征运动目标的低阶视觉运动线索,并提取目标的二值化视觉信息;突触后网络借助生物视觉系统中的稀疏不变响应特性,利用目标轮廓在连续改变形状后较大凹凸区域之间的位置关系不变特性,从低阶运动线索中编码平稳变化的视觉特征以构建行人不变表征。实验结果表明,MPDNN在公共数据集CUHK Avenue与EPFL上达到了96.96%的跨域检测准确率,比SOTA(State Of The Art)模型高4.52个百分点;在尺度、运动姿势变化数据集上也表现了较好的鲁棒性,准确率分别达到了89.48%与91.45%。以上实验结果验证了生物不变性物体识别机制在运动行人检测中的有效性。展开更多
文摘A method for moving object recognition and tracking in the intelligent traffic monitoring system is presented. For the shortcomings and deficiencies of the frame-subtraction method, a redundant discrete wavelet transform （RDWT） based moving object recognition algorithm is put forward, which directly detects moving objects in the redundant discrete wavelet transform domain. An improved adaptive mean-shift algorithm is used to track the moving object in the follow up frames. Experimental results show that the algorithm can effectively extract the moving object, even though the object is similar to the background, and the results are better than the traditional frame-subtraction method. The object tracking is accurate without the impact of changes in the size of the object. Therefore the algorithm has a certain practical value and prospect.
文摘Compressive sensing is a revolutionary idea proposed recently to achieve much lower sampling rate for signals.In the image application with limited resources the camera data can be stored and processed in compressed form.An algorithm for moving object and region detection in video using a compressive sampling is developed.The algorithm estimates motion information of the moving object and regions in the video from the compressive measurements of the current image and background scene.The algorithm does not perform inverse compressive operation to obtain the actual pixels of the current image nor the estimated background.This leads to a computationally efficient method and a system compared with the existing motion estimation methods.The experimental results show that the sampling rate can reduce to 25% without sacrificing performance.
基金This project was supported by the foundation of the Visual and Auditory Information Processing Laboratory of BeijingUniversity of China (0306) and the National Science Foundation of China (60374031).
文摘Moving object detection is one of the challenging problems in video monitoring systems, especially when the illumination changes and shadow exists. Amethod for real-time moving object detection is described. Anew background model is proposed to handle the illumination varition problem. With optical flow technology and background subtraction, a moving object is extracted quickly and accurately. An effective shadow elimination algorithm based on color features is used to refine the moving obj ects. Experimental results demonstrate that the proposed method can update the background exactly and quickly along with the varition of illumination, and the shadow can be eliminated effectively. The proposed algorithm is a real-time one which the foundation for further object recognition and understanding of video mum＇toting systems.
基金Project(40971219)supported by the Natural Science Foundation of ChinaProjects(201121202020005,T201221207)supported by the Fundamental Research Fund for the Central Universities,China
文摘A novel cast shadow detection approach was proposed.A stereo vision system was used to capture images instead of traditional single camera.It was based on an assumption that cast shadows were on a special plane.The image obtained from one camera was inversely projected to the plane and then transformed to the view from another camera.The points on the plane shared the same position between original image and the transformed image.As a result,the cast shadows can be detected.In order to improve the efficiency of cast shadow detection and decrease computational complexity,the obvious object areas in CIELAB color space were removed and the potential shadow areas were obtained.Experimental results demonstrate that the proposed approach can detect cast shadows accurately even under various illuminations.
基金Project(T201221207)supported by the Fundamental Research Fund for the Central Universities,ChinaProject(2012CB725301)supported by National Basic Research and Development Program,China
文摘A novel moving object detection method was proposed in order to adapt the difficulties caused by intermittent object motion,thermal and dynamic background sequences.Two groups of complementary Gaussian mixture models were used.The ghost and real static object could be classified by comparing the similarity of the edge images further.In each group,the multi resolution Gaussian mixture models were used and dual thresholds were applied in every resolution in order to get a complete object mask without much noise.The computational color model was also used to depress illustration variations and light shadows.The proposed method was verified by the public test sequences provided by the IEEE Change Detection Workshop and compared with three state-of-the-art methods.Experimental results demonstrate that the proposed method is better than others for all of the evaluation parameters in intermittent object motion sequences.Four and two in the seven evaluation parameters are better than the others in thermal and dynamic background sequences,respectively.The proposed method shows a relatively good performance,especially for the intermittent object motion sequences.
文摘近年来,随着羊只养殖向大规模和精细化的方向发展,羊场对智能化管理的需求日益增加。因此,精准的个体识别和行为监测变得尤为重要,对多目标跟踪(Multiple object tracking, MOT)算法的准确性提出了更高要求。然而,现有的MOT算法在目标遮挡和动态场景下的性能仍不理想。本文提出两种跟踪线索:深度调制交并比(Depth modulated intersection over union, DIoU)和轨迹方向建模(Tracklet direction modeling, TDM),旨在补充交并比(Intersection over union, IoU)线索,提高多目标跟踪的精准度和鲁棒性。DIoU线索通过引入目标的深度信息改进了传统的IoU计算方法。TDM聚焦于目标的运动趋势,预测其未来的移动方向。本文将DIoU和TDM跟踪线索集成到BoT-SORT算法中,形成改进的多目标跟踪算法。在两个私有数据集上,改进算法相比基线方法,MOTA(Multiple object tracking accuracy)指标分别提高1.6、1.7个百分点,IDF1(Identification F1 score)指标分别提高1.9、1.0个百分点。结果显示,改进算法在复杂场景中的跟踪连续性和准确性显著提升。
文摘行人作为非刚性物体,对它的视觉特征进行有效的不变表示是提高识别效果的关键。在自然视觉场景中,运动行人通常会发生尺度、背景、姿态等变化,这对用现有技术提取这些不规则特征造成阻碍。针对该问题,基于哺乳动物视网膜神经结构特性,探究运动行人不变性识别问题,并提出一种适用于视觉场景的运动行人检测神经网络(MPDNN)。MPDNN包括2个神经模块:突触前网络和突触后网络。其中,突触前网络感知表征运动目标的低阶视觉运动线索,并提取目标的二值化视觉信息;突触后网络借助生物视觉系统中的稀疏不变响应特性,利用目标轮廓在连续改变形状后较大凹凸区域之间的位置关系不变特性,从低阶运动线索中编码平稳变化的视觉特征以构建行人不变表征。实验结果表明,MPDNN在公共数据集CUHK Avenue与EPFL上达到了96.96%的跨域检测准确率,比SOTA(State Of The Art)模型高4.52个百分点;在尺度、运动姿势变化数据集上也表现了较好的鲁棒性,准确率分别达到了89.48%与91.45%。以上实验结果验证了生物不变性物体识别机制在运动行人检测中的有效性。
