No abstract

Internet of things (IoT) is a ubiquitous network that helps the system to monitor and organize the world through processing, collecting, and analyzing the data produced by IoT objects. The accurate localization of IoT objects is indispensable for most IoT applications, especially healthcare monitoring. Utilizing GPS as the positioning system is not cost‐efficient and does not apply to some environments (e.g., deep forests, oceans, inside the buildings, etc.). Hereupon, copious position estimation approaches are developed in the literature. Among range‐free approaches, distance vector‐Hop (DV‐Hop) is the widely used algorithm due to its straightforward applicability and can estimate the position of unknown objects that are far‐off the anchors. Due to its low accuracy, various techniques were proposed to increase the accuracy of basic DV‐Hop. In the most recent approach, meta‐heuristic algorithms were used, the results of which were promising. In the present paper, Tunicate Swarm Algorithm and Harris hawk optimization were initially hybridized. Afterthought, the resulting hybrid algorithm was enhanced by appending a new phase. Then, the proposed hybrid algorithm was intermingled with the DV‐Hop algorithm. In the first set of experiments, the proposed hybrid algorithm was evaluated on 50 test functions using average, SD, box plot, and p‐value criteria. In the second part, the proposed localization algorithm's efficiency was investigated in twenty‐eight different manners using node localization error, average localization error, and localization error variance metrics. The effectiveness of the contributions was evident from the experimental results.

No abstract

Recently, convolutional neural networks have shown promising performance for single-image super-resolution. In this paper, we propose Deep Artifact-Free Residual (DAFR) network which uses the merits of both residual learning and usage of ground-truth image as target. Our framework uses a deep model to extract the high frequency information which are necessary for high quality image reconstruction. We use a skip-connection to feed the low-resolution image to network before the image reconstruction. In this way, we are able to use the ground-truth images as target and avoid misleading the network due to artifacts in difference image. In order to extract clean high frequency information, we train network in two steps. First step is a traditional residual learning which uses the difference image as target. Then, the trained parameters of this step are transferred to the main training in second step. Our experimental results show that the proposed method achieves better quantitative and qualitative image quality compared to the existing methods.

No abstract