A visual sensor network for object recognition: Testbed realization

Redondi, Alessandro E. C.; Baroffio, Luca; Canclini, Antonio; Cesana, Matteo; Tagliasacchi, Marco

doi:10.1109/icdsp.2013.6622789

Cited by 16 publications

(15 citation statements)

References 20 publications

(18 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For BRISK we consider 4 octave and 4 intra-octave layers. We performed the interest point detection and feature extraction using BRISK and using SURF on a desktop computer, but SURF and BRISK implementations for battery powered sensor platforms with limited memory exist [16], [25], and since the algorithms are deterministic, they would produce the same results on a VSN node.…”

Section: Evaluation Methodologymentioning

confidence: 99%

“…As demonstrated in [20], [25], the time needed for interest point detection and for feature extraction are linear functions of the image size (in pixels) and of the number of interest points found in the image, both for SURF and for BRISK, on various hardware platforms. Furthermore, since SURF and BRISK descriptors have a fixed size, the number of interest points also influences the amount of data to be transmitted to the server node .…”

Section: Vsn Workload: Number Of Interest Pointsmentioning

confidence: 99%

See 1 more Smart Citation

Characterization of SURF and BRISK Interest Point Distribution for Distributed Feature Extraction in Visual Sensor Networks

Dán

Khan

Fodor

2015

IEEE Trans. Multimedia

View full text Add to dashboard Cite

We study the statistical characteristics of SURF and BRISK interest points and descriptors, with the aim of supporting the design of distributed processing across sensor nodes in a resource-constrained visual sensor network (VSN). Our results show high variability in the density, the spatial distribution, and the octave layer distribution of the interest points. The high variability implies that balancing the processing load among the sensor nodes is a very challenging task, and obtaining a priori information is essential, e.g., through prediction. Our results show that if a priori information is available about the images, then Topinterest point selection, limited, octave-based processing at the camera node, together with area-based interest point detection and extraction at the processing nodes, can balance the processing load and limit the transmission cost in the network. Complete interest point detection at the camera node with optimized descriptor extraction delegation to the processing nodes in turn can further decrease the transmission load and allow a better balance of the processing load among the network nodes. Index Terms-BRISK, distributed feature extraction, interest point distribution, SURF, visual sensor network (VSN).1520-9210

show abstract

Section: Evaluation Methodologymentioning

confidence: 99%

Section: Vsn Workload: Number Of Interest Pointsmentioning

confidence: 99%

Characterization of SURF and BRISK Interest Point Distribution for Distributed Feature Extraction in Visual Sensor Networks

Dán

Khan

Fodor

2015

IEEE Trans. Multimedia

View full text Add to dashboard Cite

show abstract

“…of interest points detected; this model was validated recently in [1,2]. We can thus model the detection time for slice v from sensor s at processing node n as a function of the image slice width y i s,v and of the number ξ i s,v of interest points detected in the image slice as an affine function P n (y i s,v + α f ξ i s,v ), where P n is the per unit processing time of node n. Note that ξ i s,v is unknown before processing image slice v, but efficient low-complexity predictors exist, such as the last value predictor [2].…”

Section: Visual Feature Extractionmentioning

confidence: 97%

“…As shown in [1], significant delay is incurred both when processing is performed at the sensor nodes, and when images are transmitted across the network to a central processing node. Transmitting the images across the network may also drain the energy resources of nodes relaying the images.…”

Section: Introductionmentioning

confidence: 99%

Efficient distribution of visual processing tasks in multi-camera visual sensor networks

Eriksson

Pacifici

Dán

2015

2015 IEEE International Conference on Multimedia &Amp; Expo Workshops (ICMEW)

View full text Add to dashboard Cite

Multi-camera visual sensor networks (VSNs) require large computational resources in order to perform visual analysis in real-time. One way to match the computational needs is to augment the VSN with dedicated processing nodes that do innetwork processing, but this requires careful allocation of loads from the sensor nodes in order to ensure low processing times. In this paper we formulate the problem of load allocation and completion time minimization in a VSN as an optimization problem. We propose a distributed algorithm for load allocation, and evaluate its performance in terms of completion time and convergence compared to a Greedy algorithm. Simulations show that the proposed algorithm converges faster, but at the cost of increased completion times. Nonetheless, combined with appropriate coordination, the proposed algorithm achieves low completion times at low complexity.

show abstract

“…In particular, * Matteo Cesana is also with MobiMesh srl (Spin-off Politecnico di Milano) extracting features from visual data is often a computationallyintensive task [1]. For the case of local features, this process entails detecting image keypoints and computing the corresponding descriptors, a process whose computational complexity grows linearly with the image size and with the number of scales of the image which are processed 1 [2].…”

Section: Introductionmentioning

confidence: 99%

A Mathematical Programming Approach to Task Offloading in Visual Sensor Networks

Redondi

Cesana

Baroffio

et al. 2015

2015 IEEE 81st Vehicular Technology Conference (VTC Spring)

View full text Add to dashboard Cite

Abstract-This work studies how visual analysis tasks based on feature extraction can be speeded up in the context of Visual Sensor Networks. The main catch is for the camera node to leverage the presence of neighboring sensor nodes and offload the task, thus parallelizing its execution. We propose two mathematical programming formulations for the optimal visual task offloading problem: the first one targets the minimization of the overall task completion time while enforcing energy consumption constraints onto the nodes; the second maximizes the overall sensor network lifetime subject to a temporal constraint on the task completion time. The aforementioned formulations are used to characterize the achievable speed-up and consequent energy consumption in representative visual sensor network topologies.

show abstract

A visual sensor network for object recognition: Testbed realization

Cited by 16 publications

References 20 publications

Characterization of SURF and BRISK Interest Point Distribution for Distributed Feature Extraction in Visual Sensor Networks

Characterization of SURF and BRISK Interest Point Distribution for Distributed Feature Extraction in Visual Sensor Networks

Efficient distribution of visual processing tasks in multi-camera visual sensor networks

A Mathematical Programming Approach to Task Offloading in Visual Sensor Networks

Contact Info

Product

Resources

About