Deep Learning Semantic Segmentation Research Articles

High resolution remote sensing recognition of elm sparse forest via deep-learning-based semantic segmentation

Elm (Ulmus pumila L.) sparse forest plays an vital role in maintaining local ecological stability and security in the Otingdag Sandy Land area. Prior studies on elm canopy extraction have predominantly relied on manual parameter configuration, resulting in unsatisfactory levels of generalization. To meet the needs of high-precision and rapid recognition of elm sparse forests in large areas, this study proposed a recognition method for elm sparse forest that orients to high spatial resolution remote sensing imageries, using deep-learning-based semantic segmentation techniques. It can automatically learn features that are conducive to segmenting the canopy of elm trees, and retains good generalization ability on the Gaofen-2 imageries obtained in different regions. First, we constructed a dataset specialized for elm canopy semantic segmentation task, and annotated over 130,000 elm canopies based on Gaofen-2 imageries. In addition, we trained 7 deep-learning semantic segmentation model candidates. Among them, MANet showed the best performance, with its F1-score reaching 81.44%. Lastly, we applied edge detection to the elm canopy coverage area, and automatically extract the elm canopy. The proposed method can provide technical support for the investigation and monitoring of elm sparse forests, while facilitates local desertification prevention efforts in the entire Otingdag Sandy Region.

Greener view, safer drive: Using repeated field experiments and deep transfer learning technique to investigate impacts of urban road landscapes on driving performance

Driving in urban environments is an essential part of urban residents’ daily life. We still know little about impacts of a wide range of greenness on driving performance in real urban environments, after controlling for socioeconomic, demographic, driving record, and other environmental factors. This missing knowledge prevents policymakers and professionals from using appropriate planning and design of green landscapes to create safe driving environments for numerous urban residents. This study aimed to address this significant knowledge gap by using real-world driving experiments. Each of thirty-four residents performed seven driving tasks so 238 driving tasks were completed in total. Each task lasted one hour and followed a randomly assigned sequence. Road greenness and other environmental characteristics were analyzed using deep transfer learning semantic segmentation based on live videos (30 frames per second), recorded by a camera positioned to capture the driver’s eye view. A serial communication technology, known as Controller Area Network bus (CANbus), was employed to continuously measure driving performance using four parameters. A series of hierarchical regression analyses yielded three major findings: First, an increased mean of greenness was associated with improved driving performance, as demonstrated by all four parameters. Second, an increased variation of greenness was also associated with better driving performance in three parameters. Finally, the mean of greenness displayed a stronger positive relationship with driving performance than the variation of greenness in three parameters. The findings imply that both the quantity and quality of green landscapes are critical for promoting driving performance in urban areas.

BAFormer: A Novel Boundary-Aware Compensation UNet-like Transformer for High-Resolution Cropland Extraction

Utilizing deep learning for semantic segmentation of cropland from remote sensing imagery has become a crucial technique in land surveys. Cropland is highly heterogeneous and fragmented, and existing methods often suffer from inaccurate boundary segmentation. This paper introduces a UNet-like boundary-aware compensation model (BAFormer). Cropland boundaries typically exhibit rapid transformations in pixel values and texture features, often appearing as high-frequency features in remote sensing images. To enhance the recognition of these high-frequency features as represented by cropland boundaries, the proposed BAFormer integrates a Feature Adaptive Mixer (FAM) and develops a Depthwise Large Kernel Multi-Layer Perceptron model (DWLK-MLP) to enrich the global and local cropland boundaries features separately. Specifically, FAM enhances the boundary-aware method by adaptively acquiring high-frequency features through convolution and self-attention advantages, while DWLK-MLP further supplements boundary position information using a large receptive field. The efficacy of BAFormer has been evaluated on datasets including Vaihingen, Potsdam, LoveDA, and Mapcup. It demonstrates high performance, achieving mIoU scores of 84.5%, 87.3%, 53.5%, and 83.1% on these datasets, respectively. Notably, BAFormer-T (lightweight model) surpasses other lightweight models on the Vaihingen dataset with scores of 91.3% F1 and 84.1% mIoU.

Assessment of Pelvic Tilt in Anteroposterior Radiographs by Area Ratio Based on Deep Learning.

Diagnostics. Based on deep learning semantic segmentation model, we sought to assess pelvic tilt by area ratio of the lesser pelvic and the obturator foramen in anteroposterior (AP) radiographs. Pelvic tilt is a critical factor in hip and spinal surgery, commonly evaluated by medical professionals through sagittal pelvic radiographs. The inherent pelvic asymmetry, as well as potential obstructions from clothing and musculature in roentgenography, may result in ghosting and blurring artifacts, thereby complicating precise measurement. PT directly affects the area ratio of the lesser pelvis to the obturator foramen in AP radiographs. An exponential regression analysis of simulated radiographs from ten male and ten female pelvises in specific tilt positions derived a formula correlating this area ratio with PT. Two blinded investigators evaluated this formula using 161 simulated AP pelvic radiographs. A deep learning semantic segmentation model was then fine-tuned to automatically calculate the area ratio, enabling intelligent PT evaluation. This model and the regression function were integrated for automated PT measurement and tested on a dataset of 231 clinical cases. We observed no disparity between males and females in the aforementioned area ratio. The test results from two blinded investigators analyzing 161 simulated radiographs revealed a mean absolute error of 0.19° (SD±4.71°), with a correlation coefficient between them reaching 0.96. Additionally, the mean absolute error obtained from testing 231 clinical AP radiographs using the fine-tuned semantic segmentation model mentioned earlier is -0.58° (SD±5.97°). We found that using deep learning neural networks enabled a more accurate and robust automatic measurement of PT through the area ratio of the lesser pelvis and obturator foramen.

Automated structural repair based on continuous carbon fiber reinforced plastic 3D printing and online model reconstruction

AbstractAdditive manufacturing of continuous carbon fiber reinforced plastic (CCFRP) has gained increasing prominence in structural repair through material accumulation. However, the repairing process entails laborious preliminary tasks, such as positioning, model reconstruction, and data input. In this article, computer vision is integrated into a CCFRP 3D printing device and a printing path‐selecting strategy is proposed to automate the repair process with accuracy and efficiency. An irregularly damaged structure is placed arbitrarily on the printing platform. Subsequently, the damaged area is extracted by semantic segmentation to construct a 3D model. The printing paths including zig‐zag and contour offset infilling for an irregular damage area are selected, according to the prediction of flexural modulus based on the finite element method. Finally, an in‐situ composite printing is conducted based on coaxial extrusion principle. The results show that the intersection over union (IOU) of regular damaged areas by the applied pyramid scene parsing network (PSPNet) segmentation model is 0.81. Semantic segmentation is more robust than conventional feature extraction based on filtering methods. The appearance of the repaired structure is consistent with that of the surrounding structure. The combination of CV and additive manufacturing technology presents a new way to simplify the repair process and improve repair effects.Highlights An automated repairing system for irregular damaged areas is established. Deep learning for semantic segmentation is integrated into 3D printing. Finite element submodule is introduced to select optimal printing paths. The CCFRP 3D printing is introduced to surface defect repairing.

Automatic extraction and 3D modeling of real road scenes using UAV imagery and deep learning semantic segmentation

ABSTRACT The extraction of roads from UAV images is challenged by lighting, noise, occlusions, and similar non-road objects, making high-quality road extraction difficult. To addressing these issues, this study proposes an enhanced U-Net network to automate the extraction and 3D modeling of real road scenes using UAV imagery. Initially, a cascaded atrous spatial pyramid module was integrated into the encoder to capitalize on global context information, thereby refining the fuzzy segmentation outcomes. Subsequently, a module for augmenting road feature extraction was added within the channel, and a spatial attention mechanism was introduced in the decoder to enhance edge clarity. Experimental results demonstrated that this model captures more road information compared to mainstream networks and effectively incorporates topological structure perception for road extraction in complex scenarios, thus improving road connectivity. The model achieved an F1 score and mean Intersection over Union (mIoU) of 85.6% and 81.2%, respectively, on UAV images of road scenes – marking improvements of 3.9% and 3.4% over the traditional U-Net model, thereby exhibiting superior automatic road extraction capabilities. Ultimately, the model facilitated refined modeling and visual analysis of road scenes, achieving high overall accuracy and detailed local restoration of the actual scene.

ADM-SLAM: Accurate and Fast Dynamic Visual SLAM with Adaptive Feature Point Extraction, Deeplabv3pro, and Multi-View Geometry.

Visual Simultaneous Localization and Mapping (V-SLAM) plays a crucial role in the development of intelligent robotics and autonomous navigation systems. However, it still faces significant challenges in handling highly dynamic environments. The prevalent method currently used for dynamic object recognition in the environment is deep learning. However, models such as Yolov5 and Mask R-CNN require significant computational resources, which limits their potential in real-time applications due to hardware and time constraints. To overcome this limitation, this paper proposes ADM-SLAM, a visual SLAM system designed for dynamic environments that builds upon the ORB-SLAM2. This system integrates efficient adaptive feature point homogenization extraction, lightweight deep learning semantic segmentation based on an improved DeepLabv3, and multi-view geometric segmentation. It optimizes keyframe extraction, segments potential dynamic objects using contextual information with the semantic segmentation network, and detects the motion states of dynamic objects using multi-view geometric methods, thereby eliminating dynamic interference points. The results indicate that ADM-SLAM outperforms ORB-SLAM2 in dynamic environments, especially in high-dynamic scenes, where it achieves up to a 97% reduction in Absolute Trajectory Error (ATE). In various highly dynamic test sequences, ADM-SLAM outperforms DS-SLAM and DynaSLAM in terms of real-time performance and accuracy, proving its excellent adaptability.

Building point cloud reconstruction in TomoSAR based on deep learning semantic segmentation

AbstractTomographic synthetic aperture radar (TomoSAR) possesses 3D imaging capability, making it significant for building reconstruction using TomoSAR data. The reconstruction algorithm is closely related to building point cloud detection, while traditional detection methods suffer from low automation and reliance on manual configuration. This study proposes a building point cloud reconstruction method based on deep learning semantic segmentation. Initially, deep learning method is employed for end‐to‐end building point cloud segmentation, followed by point cloud reconstruction based on the segmentation results. The proposed method is simple and efficient, elevating the level of automation in point cloud processing. Experimental validation on real TomoSAR data confirms that the proposed method achieves automated and refined reconstruction of building point clouds.

ABANet: Attention boundary‐aware network for image segmentation

AbstractDeep learning techniques have attained substantial progress in various face‐related tasks, such as face recognition, face inpainting, and facial expression recognition. To prevent infection or the spread of the virus, wearing of masks in public places has been mandated following the COVID‐19 epidemic, which has led to face occlusion and posed significant challenges for face recognition systems. Most prominent masked face recognition solutions rely on mask segmentation tasks. Therefore, segmentation can be used to mitigate the negative impacts of wearing a mask and improve recognition accuracy. Mask region segmentation suffers from two main problems: there is no standard type of masks that people wear, they come in different colours and designs, and there is no publicly available masked face dataset with appropriate ground truth for the mask region. In order to address these issues, we propose an encoder–decoder framework that utilizes a boundary‐aware attention network combined with a new hybrid loss to provide a map, patch, and pixel‐level supervision. We also introduce a dataset called MFSD, with 11,601 images and 12,758 masked faces for masked face segmentation. Furthermore, we compare the performance of different cutting‐edge deep learning semantic segmentation models on the presented dataset. Experimental results on the MSFD dataset reveal that the suggested approach outperforms state‐of‐the‐art, algorithms with 97.623% accuracy, 93.814% IoU, and 96.817% F1‐score rate. Our dataset of masked faces with mask region labels and source code will be available online.

Deep Learning based Semantic Segmentation for Buildings Detection from Remote Sensing Images

Building extraction from remote sensing images is the process of automatically identifying and extracting the boundaries of buildings from high-resolution aerial or satellite images. The extracted building footprints can be used for a variety of applications, such as urban planning, disaster management, city development, land management, environmental monitoring, and 3D modeling. The results of building extraction from remote sensing images depend on several factors, such as the quality and resolution of the image and the choice of algorithm.The process of building extraction from remote sensing images typically involves a series of steps, including image pre-processing, feature extraction, and classification. Building extraction from remote sensing images can be challenging due to factors such as varying building sizes and shapes, shadows, and occlusions. However, recent advances in deep learning and computer vision techniques have led to significant improvements in the accuracy and efficiency of building extraction methods. This research presents a deep learning semantic segmentation architecture-based model for developing building detection from high resolution remote sensing images. The open-source Massachusetts dataset is used to train the suggested UNet architecture. The model is optimized using the RMSProp algorithm with a learning rate of 0.0001 for 100 epochs. After 1.52 hours of training on Google Colab the model achieved an 83.55% F1 score, which indicates strong precision and recall.

SEM Image Processing Assisted by Deep Learning to Quantify Mesoporous γ-Alumina Spatial Heterogeneity and Its Predicted Impact on Mass Transfer.

The pore network architecture of porous heterogeneous catalyst supports has a significant effect on the kinetics of mass transfer occurring within them. Therefore, characterizing and understanding structure-transport relationships is essential to guide new designs of heterogeneous catalysts with higher activity and selectivity and superior resistance to deactivation. This study combines classical characterization via N2 adsorption and desorption and mercury porosimetry with advanced scanning electron microscopy (SEM) imaging and processing approaches to quantify the spatial heterogeneity of γ-alumina (γ-Al2O3), a catalyst support of great industrial relevance. Based on this, a model is proposed for the spatial organization of γ-Al2O3, containing alumina inclusions of different porosities with respect to the alumina matrix. Using original, advanced SEM image analysis techniques, including deep learning semantic segmentation and porosity measurement under gray-level calibration, the inclusion volume fraction and interphase porosity difference were identified and quantified as the key parameters that served as input for effective tortuosity factor predictions using effective medium theory (EMT)-based models. For the studied aluminas, spatial porosity heterogeneity impact on the effective tortuosity factor was found to be negligible, yet it was proven to become significant for an inclusion content of at least 30% and an interphase porosity difference of over 20%. The proposed methodology based on machine-learning-supported image analysis, in conjunction with other analytical techniques, is a general platform that should have a broader impact on porous materials characterization.

Online sorting of surface defective walnuts based on deep learning

To address concerns regarding walnut shell damage and inadequate sorting precision during the mechanized sorting of walnuts, a walnut automatic sorting machine was designed based on deep learning and experimental research. Initially, the rationality of the design was verified through experiment. Then, three deep learning semantic segmentation algorithms, namely PSPnet, U-net, and Deeplabv3+, were selected to train walnut detection models. Results indicated that the U-net algorithm proved to be the most effective, achieving a Mean Intersection over Union of 96.71% and a Mean Pixel Accuracy value of 98.52%. Finally, performance tests were conducted on the prototype machine, yielding results with an average sorting efficiency of 51.70 kg/h, an average loss rate of 6.50%, and an average accuracy of sorting walnuts of 92.98%. The findings can provide insights for future structural improvements and operational parameter optimization of walnut automatic sorting machines.

Leveraging Deep Learning for Automated Reconstruction of Indoor Unstructured Elements in Scan-to-BIM

Abstract. Achieving automatic 3D reconstruction for indoor scenes is extremely useful in the field of scene understanding. Building information modeling (BIM) models are essential for lowering project costs, assisting in building planning and renovations, as well as improving building management efficiency. However, nearly all current available scan-to-BIM approaches employ manual or semi-automatic methods. These approaches concentrate solely on significant structured objects, neglecting other unstructured elements such as furniture. The limitation arises from challenges in modeling incomplete point clouds of obstructed objects and capturing indoor scene details. Therefore, this research introduces an innovative and effective reconstruction framework based on deep learning semantic segmentation and model-driven techniques to address these limitations. The proposed framework utilizes wall segment recognition, feature extraction, opening detection, and automatic modeling to reconstruct 3D structured models of point clouds with different room layouts in both Manhattan and non-Manhattan architectures. Moreover, it provides 3D BIM models of actual unstructured elements by detecting objects, completing point clouds, establishing bounding boxes, determining type and orientation, and automatically generating 3D BIM models with a parametric algorithm implemented into the Revit software. We evaluated this framework using publicly available and locally generated point cloud datasets with varying furniture combinations and layout complexity. The results demonstrate the proposed framework's efficiency in reconstructing structured indoor elements, exhibiting completeness and geometric accuracy, and achieving precision and recall values greater than 98%. Furthermore, the generated unstructured 3D BIM models keep essential real-scene characteristics such as geometry, spatial locations, numerical aspects, various shapes, and orientations compared to literature methods.

Hyperparameter tuning for deep learning semantic image segmentation of micro computed tomography scanned fiber-reinforced composites

Image segmentation with deep learning models has significantly improved the accuracy of the pixel-wise labeling of scientific imaging which is critical for many quantitative image analyses. This has been feasible through U-Net and related architecture convolutional neural network models. Although the adoption of these models has been widespread, their training data pool and hyperparameters have been mostly determined by educated guesses through trial and error. In this study, we present observations of how training data volume, data augmentation, and patch size affect deep learning performance within a limited data set. Here we study U-Net model training on four different samples of x-ray CT images of fiber-reinforced composites. Because the training process is not deterministic, we relied on seven-fold replication of each experimental condition to avoid under-sampling and observe model training variance. Unsurprisingly, we find greater training data volume strongly benefits individual models’ final accuracy and learning speed while depressing variance among replicates. Importantly, data augmentation has a profound benefit to model performance, especially in cases with a low abundance of ground truth, and we conclude that high coefficients of data augmentation should be used in scientific imaging semantic segmentation models. Future work to describe and measure image complexity is warranted and likely to ultimately guide researchers on the minimum required training data volume for particular scientific imaging deep learning tasks.

Detection of bladder cancer cells using quantitative interferometric label-free imaging flow cytometry.

Bladder cancer is one of the most common cancers with a high recurrence rate. Patients undergo mandatory yearly scrutinies, including cystoscopies, which makes bladder cancer highly distressing and costly. Here, we aim to develop a non-invasive, label-free method for the detection of bladder cancer cells in urine samples, which is based on interferometric imaging flow cytometry. Eight urothelial carcinoma and one normal urothelial cell lines, along with red and white blood cells, imaged quantitatively without staining by an interferometric phase microscopy module while flowing in a microfluidic chip, and classified by two machine-learning algorithms, based on deep-learning semantic segmentation convolutional neural network and extreme gradient boosting. Furthermore, urine samples obtained from bladder-cancer patients and healthy volunteers were imaged, and classified by the system. We achieved accuracy and area under the curve (AUC) of 99% and 97% for the cell lines on both machine-learning algorithms. For the real urine samples, the accuracy and AUC were 96% and 96% for the deep-learning algorithm and 95% and 93% for the gradient-boosting algorithm, respectively. By combining label-free interferometric imaging flow cytometry with high-end classification algorithms, we achieved high-performance differentiation between healthy and malignant cells. The proposed technique has the potential to supplant cystoscopy in the bladder cancer surveillance and diagnosis space.

Not just more, but more diverse: Green landscapes along urban roads may significantly reduce drivers' psychophysiological fatigue

The impact of roadside greenness on driving fatigue in real urban settings has been insufficiently investigated, presenting a critical knowledge gap for researchers, policymakers, professionals, and the public. In this onsite driving experiment, 34 urban residents completed seven driving tasks on different urban road routes in a randomized order with one-day intervals. A total of 238 tasks were conducted, each lasting an hour, assessing psychophysiological, visual, and muscular fatigue. A cardiovascular activity monitor (BioHarness) continuously measured the driver's heart rate, with lower rates indicating reduced psychophysiological fatigue. Visual and muscular fatigue were self-reported using a Visual Analog Scale questionnaire administered before, at the midpoint, and after completing the driving task. Deep transfer learning semantic segmentation analyzed road landscape characteristics and traffic conditions recorded from the drivers' view. Statistical analysis demonstrated that higher mean and variation in greenness significantly predicted lower psychophysiological fatigue after adjusting for multiple covariates. These results indicate that enhancing both the quantity and diversity of green landscapes along urban roads is vital for reducing driver's psychophysiological fatigue. This study reveals that roadside landscapes in urban settings are not trivial decorations, and they should be considered an essential component of transportation infrastructure.

Semantic segmentation of remote sensing images based on dual‐channel attention mechanism

AbstractDue to the inadequate utilization of data correlation and complementarity in the feature extraction process of multimodal remote sensing images, the paper proposes a deep learning semantic segmentation algorithm based on the Dual Channel Attention Mechanism (DCAM). This algorithm uses U‐Net as the backbone, combining the RGB remote sensing image as one input channel with the Convolutional Block Attention Module to extract colour space features. Simultaneously, it utilizes near‐infrared (NIR) as another input channel with the Self‐Attention Module (SAM) to extract shape space features. Finally, by concatenating the multi‐scale attention features of the RGB remote sensing image channel and the NIR remote sensing image channel, it achieves the correlation and complementarity of contextual features between the two modal remote sensing images. Experimental results on the GID‐15 dataset demonstrate that the DCAM algorithm significantly improves the segmentation accuracy, edge segmentation quality, and object segmentation integrity for various types of targets compared to current mainstream segmentation methods.

A Lightweight Deep Learning Semantic Segmentation Model for Optical-Image-Based Post-Harvest Fruit Ripeness Analysis of Sugar Apples (Annona squamosa)

The sugar apple (Annona squamosa) is valued for its taste, nutritional richness, and versatility, making it suitable for fresh consumption and medicinal use with significant commercial potential. Widely found in the tropical Americas and Asia’s tropical or subtropical regions, it faces challenges in post-harvest ripeness assessment, predominantly reliant on manual inspection, leading to inefficiency and high labor costs. This paper explores the application of computer vision techniques in detecting ripeness levels of harvested sugar apples and proposes an improved deep learning model (ECD-DeepLabv3+) specifically designed for ripeness detection tasks. Firstly, the proposed model adopts a lightweight backbone (MobileNetV2), reducing complexity while maintaining performance through MobileNetV2′s unique design. Secondly, it incorporates the efficient channel attention (ECA) module to enhance focus on the input image and capture crucial feature information. Additionally, a Dense ASPP module is introduced, which enhances the model’s perceptual ability and expands the receptive field by stacking feature maps processed with different dilation rates. Lastly, the proposed model emphasizes the spatial information of sugar apples at different ripeness levels by the coordinate attention (CA) module. Model performance is validated using a self-made dataset of harvested optical images categorized into three ripeness levels. The proposed model (ECD-DeepLabv3+) achieves values of 89.95% for MIoU, 94.58% for MPA, 96.60% for PA, and 94.61% for MF1, respectively. Compared to the original DeepLabv3+, it greatly reduces the number of model parameters (Params) and floating-point operations (Flops) by 89.20% and 69.09%, respectively. Moreover, the proposed method could be directly applied to optical images obtained from the surface of the sugar apple, which provides a potential solution for the detection of post-harvest fruit ripeness.

Intelligent Identification and Quantitative Characterization of Pores in Shale SEM Images Based on Pore-Net Deep-Learning Network Model

Among the various shale reservoir evaluation methods, the scanning electron microscope (SEM) image method is widely used. Its image can intuitively reflect the development stage of a shale reservoir and is often used for the qualitative characterization of shale pores. However, manual image processing is inefficient and cannot quantitatively characterize pores. The semantic segmentation method of deep learning greatly improves the efficiency of image analysis and can calculate the face rate of shale SEM images to achieve quantitative characterization. In this paper, the high-maturity shale of the Longmaxi Formation in the Changning area of Yibin City, Sichuan Province, and the low-maturity shale of Beibu Gulf Basin in China are studied. Based on the Pore-net network model, the intelligent identification and quantitative characterization of pores in shale SEM images are realized. The pore-net model is improved from the U-net deep-learning network model, which improves the ability of the network model to identify pores. The results show that the pore-net model performs better than the U-net model, FCN model, DeepLab V3 + model, and traditional binarization method. The problem of low accuracy of the traditional pore identification method is solved. The porosity of SEM images of high-maturity shale calculated by the pore-net network model is between 12 and 19% deviation from the experimental data. The calculated porosity of the SEM image of the low-maturity shale has a large deviation from the experimental data, which is between 14 and 47%. Compared with the porosity results calculated by other methods, the results calculated by pore-net are closer to the true value, which proves that the porosity calculated by the pore-net network model is reliable. The deep-learning semantic image segmentation method is suitable for pore recognition of shale SEM images. The fully convolutional neural network model is used to train the manually labeled shale SEM images, which can realize the intelligent recognition and quantitative characterization of the pores in the shale SEM images. It provides a certain reference value for the evaluation of shale oil and gas reservoirs and the study of other porous materials.

Exploring Uncertainty-Based Self-Prompt for Test-Time Adaptation Semantic Segmentation in Remote Sensing Images

Test-time adaptation (TTA) has been proven to effectively improve the adaptability of deep learning semantic segmentation models facing continuous changeable scenes. However, most of the existing TTA algorithms lack an explicit exploration of domain gaps, especially those based on visual domain prompts. To address these issues, this paper proposes a self-prompt strategy based on uncertainty, guiding the model to continuously focus on regions with high uncertainty (i.e., regions with a larger domain gap). Specifically, we still use the Mean-Teacher architecture with the predicted entropy from the teacher network serving as the input to the prompt module. The prompt module processes uncertain maps and guides the student network to focus on regions with higher entropy, enabling continuous adaptation to new scenes. This is a self-prompting strategy that requires no prior knowledge and is tested on widely used benchmarks. In terms of the average performance, our method outperformed the baseline algorithm in TTA and continual TTA settings of Cityscapes-to-ACDC by 3.3% and 3.9%, respectively. Our method also outperformed the baseline algorithm by 4.1% and 3.1% on the more difficult Cityscapes-to-(Foggy and Rainy) Cityscapes setting, which also surpasses six other current TTA methods.