Large Vision Model (LVM) has recently demonstrated great potential for medical imaging tasks, potentially enabling image enhancement for sparse-view Cone-Beam Computed Tomography (CBCT), despite requiring a substantial amount of data for training. Meanwhile, Deep Image Prior (DIP) effectively guides an untrained neural network to generate high-quality CBCT images without any training data. However, the original DIP method relies on a well-defined forward model and a large-capacity backbone network, which is notoriously difficult to converge. In this paper, we propose a stable optimization method for the forward-model-free, LVM-based DIP model for sparse-view CBCT. Our approach consists of two main characteristics: (1) multi-scale perceptual loss (MSPL) which measures the similarity of perceptual features between the reference and output images at multiple resolutions without the need for any forward model, and (2) a reweighting mechanism that stabilizes the iteration trajectory of MSPL. One shot optimization is used to simultaneously and stably reweight MSPL and optimize LVM. We evaluate our approach on two publicly available datasets: SPARE and Walnut. The results show significant improvements in both image quality metrics and visualization that demonstrates reduced streak artifacts. The source code is available upon request.

A large number of Internet of Things (IoT) devices such as sensor nodes are deployed in various urban infrastructures to monitor surrounding information. However, it is still a challenging issue to collect data in a low‐cost, high‐quality, and reliable manner through IoT technique. Although the recruitment of mobile vehicles (MVs) to collect urban data has proved to be an effective method, most existing data collection systems lack a trust detection mechanism for malicious terminal nodes and malicious vehicles, which should lead to security vulnerabilities in practice. This paper proposes a novel data collection strategy based on a layered trust mechanism (DC‐LTM). The strategy recruits MVs as data collectors of the sensor nodes based on the data value in the city, evaluates the trustworthiness of the data reported by the nodes, and records the results to the cloud data center. Furthermore, in order to make the data collection system more efficient and trust mechanism more reliable, we introduce unmanned aerial vehicles (UAVs) dispatched by data centers to actively verify the core sensor node data and use the core sensor data as baseline data to evaluate the credibility of the vehicles and the trust value of the whole network sensor nodes. Different from the previous strategies, UAVs adopts the DC‐LTM method to obtain the node data while actively obtaining the trust value of MVs and nodes, which effectively improves the quality of data acquisition. Simulation results show that the mechanism effectively distinguishes malicious vehicles that provide false data in exchange for payment and reduces the total cost of system recruitment payments. At the same time, the proposed incentive mechanism encourages vehicle to complete the evaluation task and improves the accuracy of node trust evaluation. The recognition rates of false data attacks and flooding attacks as well as the recognition error rate of normal nodes are 100%, 98.9%, and 3.9%, respectively, which improves the quality of system data collection as a whole.