- 标题
- 摘要
- 关键词
- 实验方案
- 产品
-
Multiview Layer Fusion Model for Action Recognition Using RGBD Images
摘要： Vision-based action recognition encounters different challenges in practice, including recognition of the subject from any viewpoint, processing of data in real time, and offering privacy in a real-world setting. Even recognizing profile-based human actions, a subset of vision-based action recognition, is a considerable challenge in computer vision which forms the basis for an understanding of complex actions, activities, and behaviors, especially in healthcare applications and video surveillance systems. Accordingly, we introduce a novel method to construct a layer feature model for a profile-based solution that allows the fusion of features for multiview depth images. This model enables recognition from several viewpoints with low complexity at a real-time running speed of 63 fps for four profile-based actions: standing/walking, sitting, stooping, and lying. The experiment using the Northwestern-UCLA 3D dataset resulted in an average precision of 86.40%. With the i3DPost dataset, the experiment achieved an average precision of 93.00%. With the PSU multiview profile-based action dataset, a new dataset for multiple viewpoints which provides profile-based action RGBD images built by our group, we achieved an average precision of 99.31%.
关键词： privacy-preserving surveillance,layer fusion model,real-time processing,Multiview action recognition,RGBD images,depth-based features
更新于2025-09-23 15:22:29
-
[Lecture Notes in Computer Science] Pattern Recognition and Computer Vision Volume 11259 (First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part IV) || Asymmetric Two-Stream Networks for RGB-Disparity Based Object Detection
摘要： Currently, most methods of object detection are monocular-based. However, due to the sensitivity to color, these methods can not handle many hard samples. With the depth information, disparity maps are helpful to get over this problem. In this paper, we propose the asymmetric two-stream networks for RGB-Disparity based object detection. Our method consists of two networks, Disparity Representations Mining Network (DRMN) and Muti-Modal Detection Network (MMDN), to combine RGB and disparity data for more accurate detection. Unlike normal two-stream networks, our model is asymmetric because of the di?erent capacity of RGB and disparity data. We are the ?rst to propose a deep learning based framework utilizing only binocular information for object detection. The experiment results on KITTI and our proposed BPD dataset demonstrate that our method can achieve a signi?cant increase in performance e?ciently and get the state-of-the-art.
关键词： Two-stream networks,Object detection,RGBD data
更新于2025-09-23 15:21:21
-
[IEEE SoutheastCon 2018 - St. Petersburg, FL (2018.4.19-2018.4.22)] SoutheastCon 2018 - RGBD-Sphere SLAM
摘要： This article proposes a SLAM algorithm referred to as RGBD-Sphere SLAM. The key innovation of this work is the prototypical system that demonstrates how formal models of 3D geometric shape and appearance can be transformed into generative classification models that detect and recognize these shapes. Object models are specified as shape programs in PSML; a custom-built procedural language for 3D object modeling. Classifiers for each PSML shape are created by simulating how instances of each shape manifest in real-world sensor data, e.g., color images and range images. The proposed RGBD-Sphere SLAM algorithm demonstrates a prototypical example of the PSML program specifies spherical 3D objects having diffuse surface albedos and distinct color appearances. A recognizer uses PSML models of each object’s geometry and appearance to detect instances of these objects within streaming RGBD sensor data. The detected model parameters are then integrated into an RGBD SLAM algorithm; hence the name RGBD-Sphere SLAM. This article describes the PSML programs, the spherical detection and recognition algorithms used, and describes the impact this approach has for improving the performance of RGBD SLAM approaches by incorporating detected objects as landmarks. This is the first example of a prototypical system that externalizes the geometric and appearance modeling to a programming language from which a recognizer is created, and marks an important step towards enabling users to “program” their problem space and allow computers to transform the formal object models, as expressed in PSML, into customized classifiers suited for specific sensor suites, e.g., color imagery and depth imagery.
关键词： object recognition,RGBD,SLAM,3D object modeling,PSML
更新于2025-09-23 15:21:01
-
[Lecture Notes in Computer Science] Pattern Recognition and Computer Vision Volume 11256 (First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part I) || Co-saliency Detection for RGBD Images Based on Multi-constraint Superpixels Matching and Co-cellular Automata
摘要： Co-saliency detection aims at extracting the common salient regions from an image group containing two or more relevant images. It is a newly emerging topic in computer vision community. Different from the existing co-saliency methods focusing on RGB images, this paper proposes a novel co-saliency detection model for RGBD images, which utilizes the depth information to enhance identi?cation of co-saliency. First, we utilize the existing single saliency maps as the initialization, then we use multiple cues to compute combination inter-images similarity to match inter-neighbors for each superpixel. Especially, we extract high dimensional features for each image region with a deep convolutional neural network as semantic cue. Finally, we introduce a modi?ed 2-layer Co-cellular Automata to exploit depth information and the intrinsic relevance of similar regions through interactions with neighbors in multi-scene. The experiments on two RGBD co-saliency datasets demonstrate the effectiveness of our proposed framework.
关键词： Multi-constraint,Cellular automata,Semantic feature,RGBD,Co-saliency
更新于2025-09-10 09:29:36
-
Indoor Object Recognition in RGBD Images with Complex-Valued Neural Networks for Visually-Impaired People
摘要： We present a new multi-modal technique for assisting visually-impaired people in recognizing objects in public indoor environment. Unlike common methods which aim to solve the problem of multi-class object recognition in a traditional single-label strategy, a comprehensive approach is developed here allowing samples to take more than one label at a time. We jointly use appearance and depth cues, specifically RGBD images, to overcome issues of traditional vision systems using a new complex-valued representation. Inspired by complex-valued neural networks (CVNNs) and multi-label learning techniques, we propose two methods in order to associate each input RGBD image to a set of labels corresponding to the object categories recognized at once. The first one, ML-CVNN, is formalized as a ranking strategy where we make use of a fully complex-valued RBF network and extend it to be able to solve multi-label problems using an adaptive clustering method. The second method, L-CVNNs, deals with problem transformation strategy where instead of using a single network to formalize the classification problem as a ranking solution for the whole label set, we propose to construct one CVNN for each label where the predicted labels will be later aggregated to construct the resulting multi-label vector. Extensive experiments have been carried on two newly collected multi-labeled RGBD datasets prove the efficiency of the proposed techniques.
关键词： Multi-label learning,Complex-Valued Neural Networks,RGBD,Object Recognition,Visually-impaired people
更新于2025-09-10 09:29:36
-
Unified Image Fusion Framework with Learning-Based Application-Adaptive Importance Measure
摘要： This paper presents a novel unified image fusion framework based on an application-adaptive importance measure. In the proposed framework, an important area is selected using the importance measure obtained for each image type in each application. The key is to learn this application-adaptive importance measure that can select the important area irrespective of the input image type without manually designing the algorithm for each application. Then, the fused intensity is generated using Poisson image reconstruction. Experimental results demonstrate that the proposed framework is effective for various applications including depth-perceptible image enhancement, temperature-preserving image fusion, and haze removal.
关键词： RGBD,NIR,Image fusion,FIR,Image enhancement
更新于2025-09-10 09:29:36