RZRan Zhang

Ran Zhang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ran Zhang has a total of 15 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Visual and Textual Information Fusion-Based Zero-Shot Framework for Hazardous Material Placard Detection and Recognition

    by Ran Zhang, Zhila Bahrami, Ke Feng, Zheng Liu

    Automatically detecting and recognizing hazardous material placards using computer vision-based methods ensures safe operations and proper management of dangerous freight transportation. Deep learning-based object detection methods provide viable and practical solutions to varied applications. However, contemporary deep learning-based methods suffer from imbalanced and unseen classes, which are very common in real-life data. Thus, this study, drawing attention to this hitherto neglected challenge in real-world applications, proposes a deep learning-based zero-shot framework to detect and recognize the hazardous material placards of both imbalanced and open classes. A logarithmic weighted cross-entropy (LWCE) is proposed to balance the closed classes during training. In addition, a logarithmic weighted confidence fusion (LWCF) strategy is designed to fuse the separately extracted visual and textual information. The experiments on real-world transportation data demonstrated the proposed framework's effectiveness and superiority over other state-of-the-art methods. Notably, our framework outperforms the previous method with a remarkable margin of 12.8% in F1 score on the placard dataset. This study solves the imbalanced and open class problem by fusing object visual information and text information, providing a practical industrial application of the zero-shot learning concept.


  • Open Access

    An Efficient and Uncertainty-Aware Decision Support System for Disaster Response Using Aerial Imagery

    by Junchi Bin, Ran Zhang, Rui Wang, Yue Cao, Yufeng Zheng, Erik Blasch, Zheng Liu

    Efficient and robust search and rescue actions are always required when natural or technical disasters occur. Empowered by remote sensing techniques, building damage assessment can be achieved by fusing aerial images of pre- and post-disaster environments through computational models. Existing methods pay over-attention to assessment accuracy without considering model efficiency and uncertainty quantification in such a life-critical application. Thus, this article proposes an efficient and uncertain-aware decision support system (EUDSS) that evolves the recent computational models into an efficient decision support system, realizing the uncertainty during building damage assessment (BDA). Specifically, a new efficient and uncertain-aware BDA integrates the recent advances in computational models such as Fourier attention and Monte Carlo Dropout for uncertainty quantification efficiently. Meanwhile, a robust operation (RO) procedure is designed to invite experts for manual reviews if the uncertainty is high due to external factors such as cloud clutter and poor illumination. This procedure can prevent rescue teams from missing damaged houses during operations. The effectiveness of the proposed system is demonstrated on a public dataset from both quantitative and qualitative perspectives. The solution won the first place award in International Overhead Imagery Hackathon.


  • An End-to-End Framework for Shipping Container Corrosion Defect Inspection

    by Zhila Bahrami, Ran Zhang, Teng Wang, Zheng Liu

    The worldwide transportation industry relies heavily on shipping containers. Containerization has made it easier to transfer goods all over the world by guaranteeing cargo safety while in transit. To ensure the safety of goods during the transition, shipping containers should be reliable and kept in healthy conditions. Surface defect inspection of shipping containers is of great importance to guarantee the quality of containers. Customs officers must check the surface of shipping containers as they pass through terminal gates during the transition. Human visual observation is the basis for the current inspection method, which is time-consuming, labor-intensive, and possibly hazardous. The purpose of this research is to present a deep learning-based framework that can be used in conjunction with a computer vision technique to successfully and efficiently inspect corrosion defects on the surface of shipping containers. The proposed framework consists of two main modules, including High-resolution and Temporal Context region-based convolutional neural network (HRTC R-CNN) for corrosion defect detection and corrosion defect characterization (CDC) for corrosion defect inspection. HRTC R-CNN uses a multistream backbone, multidepth, and multiscale super-resolved feature generation to extract semantic information from different defect scales. The shallow network receives high-resolution images to maintain positional information, while the deep network receives low-resolution images to extract semantic information. In order to increase the framework’s performance, an attention mechanism and two memory banks are used to leverage context information from unlabeled images. In CDC, an optical flow-based image stitching is proposed to calculate the percentage of corrosion on the surface of the whole container. Experiments on the corrosion defect dataset demonstrate the excellent accuracy and resilience of our approach. This technology will expedite the process of container’s defect inspection at terminals, thereby supporting container transport logistics and supply chain processes.


  • An HRCR-CNN Framework for Automated Security Seal Detection on the Shipping Container

    by Zhila Bahrami, Ran Zhang, Rakiba Rayhana, Zheng Liu

    Shipping containers play a significant role in the global transportation industry. Containerization has improved the way cargo is transported all around the world by ensuring the security of cargo during transit. To ensure the security of shipments, security seals are applied to the containers to prevent any unauthorized entry. During the transition, customs officers need to inspect security seals when containers pass the gate of terminals. The existing inspection mechanism is based on human visual observation, which is labor-intensive, time-consuming, and potentially hazardous. This article aims to propose a deep learning-based framework with a machine vision system to effectively and efficiently inspect security seals. The proposed high-resolution with context R-CNN (HRCR-CNN) framework consists of two modules, i.e., multidepth super-resolved features generation (MDSRFG) and attention block along with the long-term memory, for seal detection. The MDSRFG reserves the multiple position information by employing multiple depth backbones. To achieve this, high-resolution images are fed to the shallow network to reserve the position information, whereas low-resolution images are fed to the deep network to extract semantic information. Attention along with long-term memory module is utilized to leverage context information of unlabeled images to improve the performance of the framework. The experimental results conducted on the security seal dataset show that our approach achieves high performance in terms of accuracy and robustness under various conditions. This technique will accelerate the process of seal detection at the terminals and thus contribute to container logistics and supply chain management.


  • A Vertical Text Spotting Model for Trailer and Container Codes

    by Ran Zhang, Zhila Bahrami, Zheng Liu

    Accurately detecting and recognizing vertical codes of trailers and containers plays a crucial role in logistic management during freight transportation, especially in the absence of horizontal codes. Existing research mainly focuses on developing text spotting models for texts in the direction from left to right, with or without rotation, in curved or straight lines. Vertical code spotting has not been fully explored yet. Thus, in this study, we propose a vertical text spotting model for vertical trailer and container codes. Our model adopts a multistage detector for better localization performance. The regional text features extracted by backbone network are pooled vertically with vertical RoIAlign to conform with the code direction for the convenience of recognition. In the recognition part, our model integrates two recognition pipelines that take cropped images and extracted features. The recognition results from two pipelines are selected by the proposed decision strategy considering uncertainties and probabilities. Experimental results show that our proposed model outperforms other state-of-the-art text spotting models in detecting and recognizing vertical codes.


  • An Adaptive Deep Learning Framework for Shipping Container Code Localization and Recognition

    by Ran Zhang, Zhila Bahrami, Teng Wang, Zheng Liu

    Shipping containers play an important role in global transportation. As container codes are the unique identifiers for shipping containers, recognizing these codes is an essential step to manage the containers and logistics. The conventional code localization methods can easily be interfered by varied noises and cannot identify the best regions for code recognition. In this paper, we propose an adaptive deep learning framework for shipping container code localization and recognition. In the framework, the noisy text regions will be removed by an adaptive score aggregation (ASA) algorithm. The code region boundaries are identified by the average to maximum suppression range (AMSR) algorithm. Thus, the predicted locations can be adjusted within this range to fit the code recognition model to achieve higher accuracy. The experimental results on the comparative study with the state-of-the-art models, including EAST, PSENet, GCRNN, and MaskTextSpotter, demonstrated the proposed framework achieved better localization performance and obtained 93.33% recognition accuracy. The processing speed reaches 1.13 fps, which is sufficient to meet the operational requirements. Thus, the proposed solution will facilitate the digital transformation of shipping container management and logistics at ports.