{"id":9,"date":"2008-10-08T16:09:25","date_gmt":"2008-10-08T08:09:25","guid":{"rendered":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/?page_id=9"},"modified":"2008-10-29T16:21:36","modified_gmt":"2008-10-29T08:21:36","slug":"multiple-camera","status":"publish","type":"page","link":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/?page_id=9","title":{"rendered":"Multiple Camera"},"content":{"rendered":"<p>More and more cameras have been applied in surveillance to monitor activities over an extended area. One problem associated with a multi-camera system is to <strong>automatically analyze<\/strong> and <strong>fuse information<\/strong> gathered from multiple cameras so that human intervention is reduced to a maximum extent.<\/p>\n<p>We assume that the problem of single camera tracking is already solved. Main problems associated with a multi-camera\u00c2\u00a0system are <strong>object representation and matching<\/strong>, <strong>topology estimation<\/strong> and <strong>data association<\/strong>.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/Englishflowchart.png\" alt=\"\" width=\"620\" height=\"270\" \/><\/p>\n<h2><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/h2>\n<h2>Experimental Setup:<\/h2>\n<p>\u00c2\u00a0<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" style=\"vertical-align: middle;\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/camerasetup.bmp\" alt=\"\" width=\"550\" height=\"284\" \/><\/p>\n<p>The experimental setup consists of three cameras with non-overlapping fields of view. The cameras are widely separated, including two outdoor settings and one indoor setting.<\/p>\n<p>Some example blobs: each column shows blobs of the same person under three disjoint views.<\/p>\n<p><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/sample.bmp\" alt=\"\" \/><\/p>\n<p>\u00c2\u00a0<\/p>\n<p>\u00c2\u00a0We also use <a href=\"http:\/\/vision.soe.ucsc.edu\/?q=node\/178\">VIPeR (Viewpoint Invariant Pedestrian Recognition)<\/a> database in our experiments. The VIPeR database contains a significant amount of viewpoint and illumination variation and contains 632 pedestrian image pairs taken from arbitrary viewpoints under varying illumination conditions.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/Viper database.png\" alt=\"\" width=\"651\" height=\"274\" \/><\/p>\n<p>\u00c2\u00a0<img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/p>\n<h2>Object representation and matching<\/h2>\n<p>Since the fields of view of multiple cameras are non-overlapping, the appearances of moving objects under these cameras may exhibit significant differences due to <strong>variations of illumination conditions, poses and camera parameters.<\/strong> Therefore, the key problem of matching\u00c2\u00a0objects across cameras is to build a representation for the appearance of objects which is robust to these variations. The appearance matching problem is similar to the query-by-example problem in CBIR.<\/p>\n<p><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/p>\n<h3>ACCV:<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/headtorsoleg.PNG\" alt=\"\" width=\"661\" height=\"312\" \/><\/p>\n<p>Diffusion distance is proposed to histogram matching to compensate for illumination changes and camera distortions.<\/p>\n<p>Related Publications:<\/p>\n<p><span style=\"font-family: Times New Roman;\"><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt;\"><strong>1. Yinghao Cai<\/strong><\/span><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt;\">, <span style=\"mso-bidi-font-weight: bold;\">Wei Chen<\/span>, <span style=\"mso-bidi-font-weight: bold;\">Kaiqi Huang<\/span>, <span style=\"mso-bidi-font-weight: bold;\">Tieniu Tan<\/span>, \u00e2\u20ac\u0153<strong><a href=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/ACCV.pdf\" target=\"_blank\"><span style=\"color: #e08105;\">Continuously Tracking Objects Across Multiple Widely Separated Cameras<\/span><\/a><\/strong>\u00e2\u20ac\u009d. The 8<sup>th<\/sup> Asian Conference on Computer Vision (1) 2007: 843-852. (ACCV <a href=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/CaiACCVposter2.ppt\" target=\"_blank\">poster<\/a>).<\/span><\/span><\/p>\n<p><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/p>\n<h3>ICIP:<\/h3>\n<p>Unlike snapshot-based appearance matching which matches objects by a single image, we focus on sequence matching to alleviate the uncertainties brought by segmentation errors and partial occlusions.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/divideblob.PNG\" alt=\"\" width=\"395\" height=\"226\" \/><\/p>\n<p>We partition the blob into regular patches for localization of color components. Each patch is represented by its dominant colors and frequencies of occurrence these colors appearing in the patch on the target.<\/p>\n<p>At the training stage, given the sequence of a queried person under one camera, the appearance model is formulated by concatenating feature vectors with the majority of votes over the sequence.<br \/>\nAt the testing stage, Bayesian inference is incorporated into the identification framework to accumulate the temporal information in the sequence.<\/p>\n<p>Related Publication:<\/p>\n<p><span style=\"font-family: Times New Roman;\">1. \u00c2\u00a0<strong><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt;\">Yinghao Cai<\/span><\/strong><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt; mso-bidi-font-weight: bold;\">, Kaiqi Huang and Tieniu Tan<\/span><span style=\"font-size: 12pt; mso-font-kerning: 0pt;\" lang=\"EN-US\">, \u00e2\u20ac\u0153<strong><a href=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/icip2008.pdf\" target=\"_blank\">Matching Tracking Sequences Across Widely Separated Cameras<\/a><\/strong>\u00e2\u20ac\u009d, accepted by <\/span><span style=\"font-size: 12pt; mso-bidi-font-style: italic;\" lang=\"EN-US\">IEEE International Conference on Image Processing, 2008. (Oral presentation)<\/span><span style=\"font-size: 12pt; mso-font-kerning: 0pt;\" lang=\"EN-US\">.<\/span><\/span><\/p>\n<h3><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/h3>\n<h3>ICPR:<\/h3>\n<p>Objects of interest, pedestrians are represented by a set of region signatures centered at points sampled from edges. The problem of frame-to-frame matching is formulated as finding corresponding points in two images as minimization of a cost function over the space of correspondence. The correspondence problem is solved under integer optimization framework where the cost function is determined by similarity of region signatures as well as geometric constraints between points.<\/p>\n<p><strong>A demo:<\/strong><\/p>\n<p><a href=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/demo_ICPR.avi\">Corresponding points in two images<\/a><\/p>\n<p>Related Publication:<\/p>\n<p>1.\u00c2\u00a0<span style=\"font-family: Times New Roman;\">\u00c2\u00a0<strong><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt;\">Yinghao Cai<\/span><\/strong><span style=\"font-size: 12pt; color: #000000; mso-font-kerning: 0pt; mso-bidi-font-weight: bold;\">, Kaiqi Huang and Tieniu Tan<\/span><span style=\"font-size: 12pt; mso-font-kerning: 0pt;\" lang=\"EN-US\">, \u00e2\u20ac\u0153<strong><a href=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/ICPR08.pdf\" target=\"_blank\"><span style=\"color: #e08105;\">Human Appearance Matching Across Multiple Non-overlapping Cameras<\/span><\/a><\/strong>\u00e2\u20ac\u009d, <span style=\"color: #000000;\">Pattern Recognition, 2008. ICPR 2008. 19th International Conference on, Tampa, USA. (Oral presentation). <\/span><\/span><\/span><\/p>\n<p><span style=\"font-family: Times New Roman;\"><span style=\"font-size: 12pt; mso-font-kerning: 0pt;\" lang=\"EN-US\"><!--more--><\/span><\/span><\/p>\n<h2><img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/h2>\n<h2>Topology estimation<\/h2>\n<p>\u00c2\u00a0<\/p>\n<p>A\u00c2\u00a0topology graph is\u00c2\u00a0an abstract network of nodes and connections as shown in Figure.\u00c2\u00a0 Why is topology graph necessary?~~~<\/p>\n<p>(1) Automatically learn a model of activity for a multi-camera surveillance network.<\/p>\n<p>(2)\u00c2\u00a0 Since moving objects do not move around the scene randomly. The topology graph provides an\u00c2\u00a0\u00c2\u00a0\u00c2\u00a0\u00c2\u00a0\u00c2\u00a0 understanding of paths that can be followed by moving objects moving within the field of view of the network of cameras.<\/p>\n<p>(3)\u00c2\u00a0 Spatial relations between cameras help to coordinate object tracking between multiple non-overlapping views.<\/p>\n<p>Publishable results are expected soon<\/p>\n<div class=\"O\">\n<div style=\"mso-line-spacing: '100 20 0'; mso-margin-left-alt: 216; mso-char-wrap: 1; mso-kinsoku-overflow: 1;\"><span style=\"font-size: 133%; font-family: \"><span style=\"left: -3.43%; color: #66ccff; font-family: Wingdings; position: absolute; mso-special-format: bullet; mso-color-index: 6;\">\u00c3\u00bc<\/span><\/span><\/div>\n<\/div>\n<p><span style=\"font-size: 24pt; font-family: \" lang=\"EN-US\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/topology.bmp\" alt=\"\" width=\"593\" height=\"477\" \/><\/span><\/p>\n<p><!--more--><\/p>\n<h2>\u00c2\u00a0<img decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/line11.gif\" alt=\"\" \/><\/h2>\n<h2>Data association:<\/h2>\n<p>\u00c2\u00a0<\/p>\n<p>The problem of data association is, given sequences as follows, how to optimally assign each object so that error rate is reduced?<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/127.0.0.1:8080\/users\/yhcai\/wp-content\/uploads\/2008\/08\/CAI_Optimalassignment.png\" alt=\"\" width=\"630\" height=\"218\" \/><\/p>\n","protected":false},"excerpt":{"rendered":"<p>More and more cameras have been applied in surveillance to monitor activities over an extended area. One problem associated with a multi-camera system is to automatically analyze and fuse information gathered from multiple cameras so that human intervention is reduced to a maximum extent. We assume that the problem of single camera tracking is already [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-9","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=\/wp\/v2\/pages\/9","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9"}],"version-history":[{"count":0,"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=\/wp\/v2\/pages\/9\/revisions"}],"wp:attachment":[{"href":"http:\/\/www.cbsr.ia.ac.cn\/users\/yhcai\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}