Compress-then-Analyze versus Analyze-then-Compress: What Is Best in Visual Sensor Networks?

Abstract

Visual Sensor Networks (VSNs) have attracted the interest of researchers worldwide in the last few years, and are expected to play a major role in the evolution of the Internet-of-Things (IoT). When used to perform visual analysis tasks, VSNs may be operated according to two different paradigms. In the traditional compress-then-analyze paradigm, images are acquired, compressed and transmitted for further analysis. Conversely, in the analyze-then-compress paradigm, image features are extracted by visual sensor nodes, encoded and then delivered to a remote destination where analysis is performed. The question this paper aims to answer is What is the best visual analysis paradigm in VSNs? To do this, first we empirically characterize the rate-energy-accuracy performance of the two aforementioned paradigms. Then, we leverage such models to formulate a resource allocation problem for VSNs. The problem optimally allocates the specific paradigm used by each camera node in the network and the related transmission source rate, with the objective of optimizing the accuracy of the visual analysis task and the VSN coverage. Experimental results over several VSNs instances demonstrate that there is no "winning" paradigm, but the best performance are obtained by allowing the coexistence of the two and by properly optimizing their utilization.

Evaluation of a duty-cycled asynchronous X-MAC protocol for vehicular sensor networks

“…According to the heavy-traffic assumption [41], each node in the network always has a packet in its buffer to be sent. Suppose a node is ready to transmit with probability p s , the probability of collision is p f , A ı defines the probability that ı of data packet arrives at node during a cycle, and A ≥ı is the probability that no less than ı data packets arrive at a node during a cycle.…”

A greedy approach for resource allocation in Virtual Sensor Networks

“…object recognition) [9]. We consider two paradigms to perform visual tasks: the Compress-Then-Analyze (CTA) and the Analyze-Then-Compress (ATC) [9], [10]. Tables II and III summarize the characteristics of the reference applications [7].…”

Energy-aware dynamic resource allocation in virtual sensor networks

“…By resorting to a fluid model, it should be ensured that all the data produced by the sensors are received by the sink nodes. This fact can be expressed using constraints (6)- (8). Constraint set (9) enforces that if a sensor node is either running an application or receiving data, then it must be active in the network.…”

“…The estimation of the power dissipation due to the processing load, which can not be neglected in multimedia applications [8], depends on factors such as the hardware architecture or the specific application implementation. Because of this, in Eq.…”

“…PERFORMANCE EVALUATION As a reference, we have focused on multimedia applications, which require the sensing, processing and delivery of multimedia content, specifically, in visual sensor networks, i.e. WSNs designed to perform visual analysis [8]. Based on the characterization of that work, the requirements vector r j = {12 kb/s, 842 KB, 69.23 MIPS} is used to represent the applications and the associated power dissipation (function f in eq.…”