[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Sign in to use this feature.

Years

Between: -

Subjects

remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline

Journals

Article Types

Countries / Regions

Search Results (25)

Search Parameters:
Keywords = histogram gradient thresholding

Order results
Result details
Results per page
Select all
Export citation of selected articles as:
22 pages, 3158 KiB  
Article
Sensitivity Analysis of Traffic Sign Recognition to Image Alteration and Training Data Size
by Arthur Rubio, Guillaume Demoor, Simon Chalmé, Nicolas Sutton-Charani and Baptiste Magnier
Information 2024, 15(10), 621; https://doi.org/10.3390/info15100621 - 10 Oct 2024
Viewed by 1131
Abstract
Accurately classifying road signs is crucial for autonomous driving due to the high stakes involved in ensuring safety and compliance. As Convolutional Neural Networks (CNNs) have largely replaced traditional Machine Learning models in this domain, the demand for substantial training data has increased. [...] Read more.
Accurately classifying road signs is crucial for autonomous driving due to the high stakes involved in ensuring safety and compliance. As Convolutional Neural Networks (CNNs) have largely replaced traditional Machine Learning models in this domain, the demand for substantial training data has increased. This study aims to compare the performance of classical Machine Learning (ML) models and Deep Learning (DL) models under varying amounts of training data, particularly focusing on altered signs to mimic real-world conditions. We evaluated three classical models: Support Vector Machine (SVM), Random Forest, and Linear Discriminant Analysis (LDA), and one Deep Learning model: Convolutional Neural Network (CNN). Using the German Traffic Sign Recognition Benchmark (GTSRB) dataset, which includes approximately 40,000 German traffic signs, we introduced digital alterations to simulate conditions such as environmental wear or vandalism. Additionally, the Histogram of Oriented Gradients (HOG) descriptor was used to assist classical models. Bayesian optimization and k-fold cross-validation were employed for model fine-tuning and performance assessment. Our findings reveal a threshold in training data beyond which accuracy plateaus. Classical models showed a linear performance decrease under increasing alteration, while CNNs, despite being more robust to alterations, did not significantly outperform classical models in overall accuracy. Ultimately, classical Machine Learning models demonstrated performance comparable to CNNs under certain conditions, suggesting that effective road sign classification can be achieved with less computationally intensive approaches. Full article
(This article belongs to the Special Issue Machine Learning and Artificial Intelligence with Applications)
Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Traffic signs samples from the GTSRB dataset.</p>
Full article ">Figure 2
<p>Illustration of a Support Vector Machine (SVM) classifier depicting the separating hyperplane (blue line), support vectors (blue dotted lines), and margin (orange lines). The two classes are represented by red and blue data points.</p>
Full article ">Figure 3
<p>Random Forest classifier. The test sample input is evaluated by multiple decision trees, each providing an individual prediction. These individual predictions are then averaged to produce the final Random Forest prediction.</p>
Full article ">Figure 4
<p>Linear Discriminant Analysis (LDA) process showing data distribution before (<b>left</b>) and after (<b>right</b>) LDA. Post-LDA, data is projected to maximize class separation, enhancing classification by preserving discriminatory information.</p>
Full article ">Figure 5
<p>ImageNet error rate trends: Since 2012, CNNs have dominated, reaching superhuman performance in image classification.</p>
Full article ">Figure 6
<p>Architecture of a Convolutional Neural Network (CNN) for image classification, illustrating the stages from input images through convolution, pooling, and fully connected layers to the final probabilistic distribution and classification output.</p>
Full article ">Figure 7
<p>Example of traffic sign alteration using the dashboard.</p>
Full article ">Figure 8
<p>Traffic signs samples from the altered GTSRB dataset.</p>
Full article ">Figure 9
<p>HOG representation (<b>right</b>) of an image from the GTSRB database (<b>left</b>).</p>
Full article ">Figure 10
<p>SVM accuracy performance as a function of the number of input images per class (blue). Threshold set at 99% of the final achieved accuracy (red).</p>
Full article ">Figure 11
<p>Evolution of model accuracy based on the number of input images for 30% alteration rate: SVM (blue), LDA (yellow), Random Forest (green), CNN (red).</p>
Full article ">Figure 12
<p>Accuracy variation with the percentage of dataset alteration for 20 images per class: SVM (blue), LDA (yellow), Random Forest (green), CNN (red).</p>
Full article ">Figure 13
<p>SVM accuracy heatmap by dataset alteration percentage and number of input images per class.</p>
Full article ">Figure 14
<p>LDA accuracy heatmap by dataset alteration percentage and number of input images per class.</p>
Full article ">Figure 15
<p>Random Forest accuracy heatmap by dataset alteration percentage and number of input images per class.</p>
Full article ">Figure 16
<p>CNN accuracy heatmap by dataset alteration percentage and number of input images per class.</p>
Full article ">
22 pages, 5638 KiB  
Article
A Method for Defogging Sea Fog Images by Integrating Dark Channel Prior with Adaptive Sky Region Segmentation
by Kongchi Hu, Qingyan Zeng, Junyan Wang, Jianqing Huang and Qi Yuan
J. Mar. Sci. Eng. 2024, 12(8), 1255; https://doi.org/10.3390/jmse12081255 - 25 Jul 2024
Viewed by 887
Abstract
Due to the detrimental impact of fog on image quality, dehazing maritime images is essential for applications such as safe maritime navigation, surveillance, environmental monitoring, and marine research. Traditional dehazing techniques, which are dependent on presupposed conditions, often fail to perform effectively, particularly [...] Read more.
Due to the detrimental impact of fog on image quality, dehazing maritime images is essential for applications such as safe maritime navigation, surveillance, environmental monitoring, and marine research. Traditional dehazing techniques, which are dependent on presupposed conditions, often fail to perform effectively, particularly when processing sky regions within marine fog images in which these conditions are not met. This study proposes an adaptive sky area segmentation dark channel prior to the marine image dehazing method. This study effectively addresses challenges associated with traditional marine image dehazing methods, improving dehazing results affected by bright targets in the sky area and mitigating the grayish appearance caused by the dark channel. This study uses the grayscale value of the region boundary’s grayscale discontinuity characteristics, takes the grayscale value with the least number of discontinuity areas in the grayscale histogram as a segmentation threshold adapted to the characteristics of the sea fog image to segment bright areas such as the sky, and then uses grayscale gradients to identify grayscale differences in different bright areas, accurately distinguishing boundaries between sky and non-sky areas. By comparing the area parameters, non-sky blocks are filled; this adaptively eliminates interference from other bright non-sky areas and accurately locks the sky area. Furthermore, this study proposes an enhanced dark channel prior approach that optimizes transmittance locally within sky areas and globally across the image. This is achieved using a transmittance optimization algorithm combined with guided filtering technology. The atmospheric light estimation is refined through iterative adjustments, ensuring consistency in brightness between the dehazed and original images. The image reconstruction employs calculated atmospheric light and transmittance values through an atmospheric scattering model. Finally, the use of gamma-correction technology ensures that images more accurately replicate natural colors and brightness levels. Experimental outcomes demonstrate substantial improvements in the contrast, color saturation, and visual clarity of marine fog images. Additionally, a set of foggy marine image data sets is developed for monitoring purposes. Compared with traditional dark channel prior dehazing techniques, this new approach significantly improves fog removal. This advancement enhances the clarity of images obtained from maritime equipment and effectively mitigates the risk of maritime transportation accidents. Full article
(This article belongs to the Section Ocean Engineering)
Show Figures

Figure 1

Figure 1
<p>A comparison of boundaries between maritime and terrestrial sky and non-sky areas. (<b>b</b>,<b>c</b>) are magnified portions of (<b>a</b>), and (<b>e</b>) is a magnified portion of (<b>d</b>). (<b>a</b>) is the maritime image. (<b>b</b>) shows the magnified boundary between the sky and maritime areas in the maritime image. (<b>c</b>) shows the magnified boundary between the sky and terrestrial areas in a maritime image. (<b>d</b>) is the terrestrial image. (<b>e</b>) shows the magnified boundary between the sky and terrestrial areas in the terrestrial image.</p>
Full article ">Figure 2
<p>Comparison of land-based defogging algorithms applied to maritime images. The input image in (<b>a</b>) is defogged using the algorithms of (<b>b</b>) Liu et al. [<a href="#B18-jmse-12-01255" class="html-bibr">18</a>], (<b>c</b>) T. M. Bui et al. [<a href="#B19-jmse-12-01255" class="html-bibr">19</a>], (<b>d</b>) He et al. [<a href="#B11-jmse-12-01255" class="html-bibr">11</a>], and (<b>e</b>) Wang et al. [<a href="#B20-jmse-12-01255" class="html-bibr">20</a>], as well as (<b>f</b>) the proposed algorithm.</p>
Full article ">Figure 3
<p>A detailed depiction of the dehazing algorithm process outlined in this study.</p>
Full article ">Figure 4
<p>Explanation of threshold selection. (<b>a</b>) Distribution of optimal segmentation thresholds for 25 foggy maritime images; (<b>b</b>) Histogram of grayscale values for the foggy maritime images.</p>
Full article ">Figure 5
<p><span class="html-italic">L</span>(i, j) represents the central pixel, with its surrounding eight pixels grouped into four sets labeled (1), (2), (3), and (4).</p>
Full article ">Figure 6
<p>Linear regression fitting depicting the relationship between the sky transmittance range and the adjustment factor γ.</p>
Full article ">Figure 7
<p>Qualitative comparison of different methods for defogging maritime images. The foggy input images (first column) were restored using the algorithms of He et al. [<a href="#B11-jmse-12-01255" class="html-bibr">11</a>] (second column), T. V. Nguyen et al. [<a href="#B22-jmse-12-01255" class="html-bibr">22</a>] (third column), Hu et al. [<a href="#B21-jmse-12-01255" class="html-bibr">21</a>] (fourth column), Kaplan N.H. et al. [<a href="#B27-jmse-12-01255" class="html-bibr">27</a>] (fifth column), and Liu et al. [<a href="#B18-jmse-12-01255" class="html-bibr">18</a>] (sixth column), as well as the proposed algorithm (seventh column). Each row, labeled from (<b>a</b>–<b>e</b>), corresponds to distinct maritime scenes, providing a side-by-side comparison of the effectiveness of each algorithm under varying fog conditions.</p>
Full article ">Figure 8
<p>Comparison of demisting outcomes for the misty maritime images and their amplified segments. The second and fourth columns correspondingly exhibit enlarged portions of the crimson rectangles in the initial and subsequent columns.</p>
Full article ">Figure 9
<p>The defogging outcomes of the proposed algorithm for varying α weights: (<b>a</b>) α = 5; (<b>b</b>) α = 25; (<b>c</b>) α = 75. The first and third rows display the outcomes of threshold-based segmentation, while the second and third columns display the corresponding defogged images.</p>
Full article ">Figure 10
<p>The defogging outcomes of the proposed algorithm for varying β weights: (<b>a</b>) β = 0.5; (<b>b</b>) β = 2.5; (<b>c</b>) β = 5; (<b>d</b>) β = 10. The first row displays the results of threshold segmentation, the second row shows the recognition of grayscale gradient outcomes, the third row illustrates connected region filling, and the fourth row presents the defogged images.</p>
Full article ">Figure 11
<p>The defogging outcomes of the proposed algorithm for varying μ weights: (<b>a</b>) μ = 0.5; (<b>b</b>) μ = 1; (<b>c</b>) μ = 5; (<b>d</b>) μ = 15; (<b>e</b>) μ = 20. The transmission maps are presented in the first row, and the corresponding defogged images are displayed in the second row.</p>
Full article ">Figure 12
<p>The defogging outcomes of the proposed algorithm for varying η weights: (<b>a</b>) η = 0; (<b>b</b>) η = 0.5; (<b>c</b>) η = 0.5; (<b>d</b>) η = 0.75; (<b>e</b>) η = 1.</p>
Full article ">Figure 13
<p>Defogging outcomes of land fog images, presenting input images in the top row and corresponding defogged results in the bottom row. (<b>a</b>,<b>b</b>) Land images without sky regions and (<b>c</b>–<b>f</b>) land images with sky regions.</p>
Full article ">
16 pages, 19129 KiB  
Article
Ship Detection in SAR Images Based on Steady CFAR Detector and Knowledge-Oriented GBDT Classifier
by Shuqi Sun and Junfeng Wang
Electronics 2024, 13(14), 2692; https://doi.org/10.3390/electronics13142692 - 10 Jul 2024
Cited by 1 | Viewed by 974
Abstract
Ship detection is a significant issue in remote sensing based on Synthetic Aperture Radar (SAR). This paper combines the advantages of a steady constant false alarm rate (CFAR) detector and a knowledge-oriented Gradient Boosting Decision Tree (GBDT) classifier to achieve the location and [...] Read more.
Ship detection is a significant issue in remote sensing based on Synthetic Aperture Radar (SAR). This paper combines the advantages of a steady constant false alarm rate (CFAR) detector and a knowledge-oriented Gradient Boosting Decision Tree (GBDT) classifier to achieve the location and the classification of ship candidates. The steady CFAR detector smooths the image by a moving-average filter and models the probability distribution of the smoothed clutter as a Gaussian distribution. The mean and the standard deviation of the Gaussian distribution are estimated according to the left half of the histogram to remove the effect of land, ships, and other targets. From the Gaussian distribution and a preset constant false alarm rate, a threshold is obtained to segment land, ships, and other targets from the clutter. Then, a series of morphological operations are introduced to eliminate land and extract ships and other targets, and an active contour algorithm is utilized to refine ships and other targets. Finally, ships are recognized from other targets by a knowledge-oriented GBDT classifier. Based on the brain-like ship-recognition process, we change the way of the decision-tree generation and achieve a higher classification performance than the original GBDT. The results on the AIRSARShip-1.0 dataset demonstrate that this scheme has a competitive performance against deep learning, especially in the detection of offshore ships. Full article
(This article belongs to the Special Issue Radar Signal Processing Technology)
Show Figures

Figure 1

Figure 1
<p>Algorithm procedure diagram of our work.</p>
Full article ">Figure 2
<p>A diagram of the CFAR detector based on the hollow shifted window.</p>
Full article ">Figure 3
<p>A flow chart of the proposed algorithm.</p>
Full article ">Figure 4
<p>A sketch of the grayscale histogram of a sea image.</p>
Full article ">Figure 5
<p>ACM data-processing pipeline.</p>
Full article ">Figure 6
<p>A typical rectangle fitting result.</p>
Full article ">Figure 7
<p>The novel regression tree structure.</p>
Full article ">Figure 8
<p>The visualization of hyperparameter selection experiments. (<b>a</b>) The visualization of <a href="#electronics-13-02692-t002" class="html-table">Table 2</a>. (<b>b</b>) The visualization of <a href="#electronics-13-02692-t003" class="html-table">Table 3</a>. (<b>c</b>) The visualization of <a href="#electronics-13-02692-t004" class="html-table">Table 4</a>. (<b>d</b>) The visualization of <a href="#electronics-13-02692-t005" class="html-table">Table 5</a>.</p>
Full article ">Figure 9
<p>The comparison of detection results of some advanced techniques and ours. (<b>a</b>) The visualized result of ground truth. (<b>b</b>) The detection result of YOLOv3. (<b>c</b>) The detection result of Faster R-CNN. (<b>d</b>) The detection result of soft teacher. (<b>e</b>) The detection result of our work.</p>
Full article ">Figure 10
<p>A detection result in the offshore scene. (<b>a</b>) The visualized result of ground truth. (<b>b</b>) The detection result of our work.</p>
Full article ">Figure 11
<p>A detection result missing the ships on the edges. (<b>a</b>) The visualized result of ground truth. (<b>b</b>) The detection result of our work.</p>
Full article ">
32 pages, 15331 KiB  
Review
Detecting Wear and Tear in Pedestrian Crossings Using Computer Vision Techniques: Approaches, Challenges, and Opportunities
by Gonçalo J. M. Rosa, João M. S. Afonso, Pedro D. Gaspar, Vasco N. G. J. Soares and João M. L. P. Caldeira
Information 2024, 15(3), 169; https://doi.org/10.3390/info15030169 - 20 Mar 2024
Cited by 1 | Viewed by 2282
Abstract
Pedestrian crossings are an essential part of the urban landscape, providing safe passage for pedestrians to cross busy streets. While some are regulated by timed signals and are marked with signs and lights, others are simply marked on the road and do not [...] Read more.
Pedestrian crossings are an essential part of the urban landscape, providing safe passage for pedestrians to cross busy streets. While some are regulated by timed signals and are marked with signs and lights, others are simply marked on the road and do not have additional infrastructure. Nevertheless, the markings undergo wear and tear due to traffic, weather, and road maintenance activities. If pedestrian crossing markings are excessively worn, drivers may not be able to see them, which creates road safety issues. This paper presents a study of computer vision techniques that can be used to identify and classify pedestrian crossings. It first introduces the related concepts. Then, it surveys related work and categorizes existing solutions, highlighting their key features, strengths, and limitations. The most promising techniques are identified and described: Convolutional Neural Networks, Histogram of Oriented Gradients, Maximally Stable Extremal Regions, Canny Edge, and thresholding methods. Their performance is evaluated and compared on a custom dataset developed for this work. Insights on open issues and research opportunities in the field are also provided. It is shown that managers responsible for road safety, in the context of a smart city, can benefit from computer vision approaches to automate the process of determining the wear and tear of pedestrian crossings. Full article
(This article belongs to the Section Wireless Technologies)
Show Figures

Figure 1

Figure 1
<p>Neural network architecture with convolutional layers.</p>
Full article ">Figure 2
<p>YOLO model detection process.</p>
Full article ">Figure 3
<p>Illustration of the architecture of the SSD model. Adapted from [<a href="#B15-information-15-00169" class="html-bibr">15</a>].</p>
Full article ">Figure 4
<p>Comparison of CNN architectures. Adapted from [<a href="#B21-information-15-00169" class="html-bibr">21</a>].</p>
Full article ">Figure 5
<p>Illustration of the architecture of the Mask R-CNN model. Adapted from [<a href="#B26-information-15-00169" class="html-bibr">26</a>].</p>
Full article ">Figure 6
<p>Example of a HOG application: (<b>a</b>) grayscale image; (<b>b</b>–<b>d</b>) selected pixel and gradient tone; (<b>e</b>) histogram graph of gradient tone. Using code available in [<a href="#B29-information-15-00169" class="html-bibr">29</a>,<a href="#B30-information-15-00169" class="html-bibr">30</a>].</p>
Full article ">Figure 7
<p>The application of HOG was used to verify the preservation of a pedestrian crossing: (<b>a</b>,<b>c</b>) original images; (<b>b</b>,<b>d</b>) images after HOG.</p>
Full article ">Figure 8
<p>An example showcasing the application of the MSER technique.</p>
Full article ">Figure 9
<p>Example of the application of the Bird’s Eye View technique: (<b>a</b>) is the original image; (<b>b</b>) is the image after application of the technique.</p>
Full article ">Figure 10
<p>Examples of the application of the Fish Eye View technique: (<b>a</b>) the original image; (<b>b</b>) the image after applying the technique.</p>
Full article ">Figure 11
<p>Examples of applying Threshold to images: Binary images are represented in (<b>a</b>,<b>b</b>) where Threshold is applied. Contour Detection is represented in (<b>c</b>,<b>d</b>).</p>
Full article ">Figure 12
<p>Illustration of images featuring less favorable elements.</p>
Full article ">Figure 13
<p>Example of edge detection using the Canny Edge algorithm.</p>
Full article ">Figure 14
<p>Illustration of the Support Vector Machine concept.</p>
Full article ">Figure 15
<p>Traditional CNN architecture compared to a MobileNet model architecture.</p>
Full article ">Figure 16
<p>Illustration of the architecture of the VGG-16 model. Adapted from [<a href="#B17-information-15-00169" class="html-bibr">17</a>].</p>
Full article ">Figure 17
<p>Cell phone setup for capturing videos.</p>
Full article ">Figure 18
<p>Routes taken to create the dataset in the city of Castelo Branco, Portugal.</p>
Full article ">Figure 19
<p>Example of creating annotations on the Roboflow platform.</p>
Full article ">Figure 20
<p>Original image (<b>a</b>) and images (<b>b</b>,<b>c</b>) after the changes have been applied.</p>
Full article ">Figure 21
<p>Method for detecting signs of usage on pedestrian crossings.</p>
Full article ">Figure 22
<p>Example of the application of HOG in the context of pedestrian crossings together with the ROI technique.</p>
Full article ">Figure 23
<p>Example of combining HOG, ROI, and perspective techniques in favorable (<b>a</b>) and unfavorable (<b>b</b>) crosswalk conditions.</p>
Full article ">Figure 24
<p>Examples of the application of image processing techniques. Different Threshold methods applied on (<b>a</b>) and (<b>b</b>), (<b>c</b>) and (<b>d</b>) shows the use of Canny Edge algorithm.</p>
Full article ">Figure 25
<p>Comparison of a fixed threshold value with the automatically calculated threshold.</p>
Full article ">Figure 26
<p>Visualization of Canny Edge and Adaptive Threshold behavior in the presence of different types of lighting, (<b>a</b>(<b>1</b>–<b>4</b>)) shows the behavior of Canny Edge and Adaptive Threshold on shadows; (<b>b</b>(<b>1</b>–<b>4</b>)) illustrates behavior on wet floor conditions.</p>
Full article ">Figure 27
<p>Comparison of Canny Edge without using a perspective and ROI, (<b>a</b>) uses Canny Edge and (<b>b</b>) uses Canny Edge and ROI.</p>
Full article ">Figure 28
<p>Application of the MSER technique, (<b>a</b>) original image; (<b>b</b>) ROI selected; (<b>c</b>,<b>d</b>) areas detected by MSER technique; (<b>e</b>) fine tunned parameters to reduce irrelevant areas.</p>
Full article ">Figure 29
<p>Illustration of the overfitting concept.</p>
Full article ">Figure 30
<p>Illustration of the early stopping solution.</p>
Full article ">Figure 31
<p>Example of a detected false positive.</p>
Full article ">Figure 32
<p>Comparison of inference time.</p>
Full article ">Figure 33
<p>Pedestrian crossing detection examples, (<b>a</b>) SSD-MobileNet-V2; (<b>b</b>) SSD-Efficient-D0; (<b>c</b>) YOLOv4-tiny.</p>
Full article ">
23 pages, 16409 KiB  
Article
Maritime Infrared Small Target Detection Based on the Appearance Stable Isotropy Measure in Heavy Sea Clutter Environments
by Fan Wang, Weixian Qian, Ye Qian, Chao Ma, He Zhang, Jiajie Wang, Minjie Wan and Kan Ren
Sensors 2023, 23(24), 9838; https://doi.org/10.3390/s23249838 - 15 Dec 2023
Cited by 2 | Viewed by 1292
Abstract
Infrared small target detection plays a crucial role in maritime security. However, detecting small targets within heavy sea clutter environments remains challenging. Existing methods often fail to deliver satisfactory performance in the presence of substantial clutter interference. This paper analyzes the spatial–temporal appearance [...] Read more.
Infrared small target detection plays a crucial role in maritime security. However, detecting small targets within heavy sea clutter environments remains challenging. Existing methods often fail to deliver satisfactory performance in the presence of substantial clutter interference. This paper analyzes the spatial–temporal appearance characteristics of small targets and sea clutter. Based on this analysis, we propose a novel detection method based on the appearance stable isotropy measure (ASIM). First, the original images are processed using the Top-Hat transformation to obtain the salient regions. Next, a preliminary threshold operation is employed to extract the candidate targets from these salient regions, forming a candidate target array image. Third, to distinguish between small targets and sea clutter, we introduce two characteristics: the gradient histogram equalization measure (GHEM) and the local optical flow consistency measure (LOFCM). GHEM evaluates the isotropy of the candidate targets by examining their gradient histogram equalization, while LOFCM assesses their appearance stability based on local optical flow consistency. To effectively combine the complementary information provided by GHEM and LOFCM, we propose ASIM as a fusion characteristic, which can effectively enhance the real target. Finally, a threshold operation is applied to determine the final targets. Experimental results demonstrate that our proposed method exhibits superior comprehensive performance compared to baseline methods. Full article
(This article belongs to the Section Optical Sensors)
Show Figures

Figure 1

Figure 1
<p>Schematic diagram of an IR image and typical salient regions. (<b>a</b>) An IR image with a small target and heavy sea clutter. (<b>b</b>) Local images of the marked regions within five consecutive frames.</p>
Full article ">Figure 2
<p>Flowchart of the proposed method.</p>
Full article ">Figure 3
<p>Structure element. The value of the effective pixels are 1, marked in red. The value of the invalid pixels are 0, marked in white.</p>
Full article ">Figure 4
<p>Intermediate results of the proposed method. (<math display="inline"><semantics> <mi mathvariant="bold">a</mi> </semantics></math>) Top-Hat image. (<math display="inline"><semantics> <mi mathvariant="bold">b</mi> </semantics></math>) Binary image. (<math display="inline"><semantics> <mi mathvariant="bold">c</mi> </semantics></math>) Candidate target array. (<math display="inline"><semantics> <mi mathvariant="bold">d</mi> </semantics></math>) Visual HOG of the candidate target array. (<math display="inline"><semantics> <mi mathvariant="bold">e</mi> </semantics></math>) Optical flow vectors of candidate target array. (<math display="inline"><semantics> <mi mathvariant="bold">f</mi> </semantics></math>) ASIM image.</p>
Full article ">Figure 5
<p>Related schematic diagrams of HOG. (<math display="inline"><semantics> <mi mathvariant="bold">a</mi> </semantics></math>) Schematic diagram of HOG calculation. The two pixels marked with blue and red circles are used as examples. (<math display="inline"><semantics> <mi mathvariant="bold">b</mi> </semantics></math>) A calculation result of HOG. (<math display="inline"><semantics> <mi mathvariant="bold">c</mi> </semantics></math>) Schematic diagram of visual HOG.</p>
Full article ">Figure 6
<p>Comparison of HOG and optical flow between small target and sea clutter. (<math display="inline"><semantics> <mi mathvariant="bold">a</mi> </semantics></math>) Visual HOG of a small target and an anisotropic clutter. (<math display="inline"><semantics> <mi mathvariant="bold">b</mi> </semantics></math>) Visual optical flow vectors of a small target and an isotropic clutter. The green arrows represent the optical flow vectors.</p>
Full article ">Figure 7
<p>Schematic diagram of characteristic neighborhood division.</p>
Full article ">Figure 8
<p>First frame of the experimental sequences. The small targets are marked with red boxes, and their enlarged views are placed in the left-bottom corner of the images.</p>
Full article ">Figure 9
<p>The resulting images of different methods on Seq.1–6. The target area in each resulting image is marked with a red box and its enlarged view is placed in the left-bottom corner.</p>
Full article ">Figure 10
<p>The resulting images of different methods on Seq.7–12. The target area in each resulting image is marked with a red box and its enlarged view is placed in the left-bottom corner.</p>
Full article ">Figure 11
<p>Receiver operating characteristic (ROC) curves of different methods.</p>
Full article ">
23 pages, 12146 KiB  
Article
Otsu Image Segmentation Algorithm Based on Hybrid Fractional-Order Butterfly Optimization
by Yu Ma, Ziqian Ding, Jing Zhang and Zhiqiang Ma
Fractal Fract. 2023, 7(12), 871; https://doi.org/10.3390/fractalfract7120871 - 8 Dec 2023
Cited by 1 | Viewed by 1576
Abstract
To solve the drawbacks of the Otsu image segmentation algorithm based on traditional butterfly optimization, such as slow convergence speed and poor segmentation accuracy, this paper proposes hybrid fractional-order butterfly optimization with the Otsu image segmentation algorithm. G-L-type fractional-order differentiation is combined with [...] Read more.
To solve the drawbacks of the Otsu image segmentation algorithm based on traditional butterfly optimization, such as slow convergence speed and poor segmentation accuracy, this paper proposes hybrid fractional-order butterfly optimization with the Otsu image segmentation algorithm. G-L-type fractional-order differentiation is combined with the algorithm’s global search to improve the position-updating method, which enhances the algorithm’s convergence speed and prevents it from falling into local optima. The sine-cosine algorithm is introduced in the local search step, and Caputo-type fractional-order differentiation is used to avoid the disadvantages of the sine-cosine algorithm and to improve the optimization accuracy of the algorithm. By dynamically converting the probability, the ratio of global search to local search is changed to attain high-efficiency and high-accuracy optimization. Based on the 2-D grayscale gradient distribution histogram, the trace of discrete matrices between classes is chosen as the fitness function, the best segmentation threshold is searched for, image segmentation is processed, and three categories of images are chosen to proceed with the segmentation test. The experimental results show that, compared with traditional butterfly optimization, the convergence rate of hybrid fractional-order butterfly optimization with the Otsu image segmentation algorithm is improved by about 73.38%; meanwhile, it has better segmentation accuracy than traditional butterfly optimization. Full article
Show Figures

Figure 1

Figure 1
<p>Planar 2−D distribution histogram. I and III represent the background and the target, and the II and IV parts represent the edge and the noise.</p>
Full article ">Figure 2
<p>Convergence curves of 5 algorithms on 12 benchmark functions.</p>
Full article ">Figure 3
<p>The flow chart of HFBOA-Otsu.</p>
Full article ">Figure 4
<p>Human image segmentation results of five algorithms. The first row is the Lena image, the second row is the Pirate image, the third row is the Woman-blonde image, and the fourth row is the Kodim image. The first column is the original images, and the second column to the sixth column are the segmentation results of the BOA-Otsu algorithm, the FFA-Otsu algorithm, the Im-Fpso Otsu algorithm, the PSO-Otsu algorithm, and the HFBOA-Otsu algorithm, respectively.</p>
Full article ">Figure 5
<p>The fitness curves of human images. (<b>a</b>) The fitness curve of the Lena image; (<b>b</b>) the fitness curve of the Pirate image; (<b>c</b>) the fitness curve of the Woman-blonde image; (<b>d</b>) the fitness curve of the Kodim image.</p>
Full article ">Figure 6
<p>Scenery images and segmentation results of five algorithms. The first row is the Wall image, the second row is the Gorge image, the third row is the Butterfly image, and the fourth row is the Mandril image. The first column is the original images, and the second column to the sixth column are the segmentation results of the BOA-Otsu algorithm, the FFA-Otsu algorithm, the Im-Fpso Otsu algorithm, the PSO-Otsu algorithm, and the HFBOA-Otsu algorithm, respectively.</p>
Full article ">Figure 7
<p>The fitness curves of scenery images. (<b>a</b>) The fitness curve of the Wall image; (<b>b</b>) the fitness curve of the Gorge image; (<b>c</b>) the fitness curve of the Butterfly image; (<b>d</b>) the fitness curve of the Mandril image.</p>
Full article ">Figure 8
<p>Medical images and segmentation results of five algorithms. The first row is the Lung1 image, the second row is the Lung2 image, the third row is the Thorax image, and the fourth row is the Brain image. The first column is the original images, and the second column to the sixth column are the segmentation results of the BOA-Otsu algorithm, the FFA-Otsu algorithm, the Im-Fpso Otsu algorithm, the PSO-Otsu algorithm, and the HFBOA-Otsu algorithm, respectively.</p>
Full article ">Figure 9
<p>The fitness curves of medical images. (<b>a</b>) The fitness curve of the Lung1 image; (<b>b</b>) the fitness curve of the Lung2 image; (<b>c</b>) the fitness curve of the Thorax image; (<b>d</b>) the fitness curve of the Brain image.</p>
Full article ">
13 pages, 16053 KiB  
Article
Analysis of the Possibility of Using Selected Tools and Algorithms in the Classification and Recognition of Type of Microstructure
by Michał Szatkowski, Dorota Wilk-Kołodziejczyk, Krzysztof Jaśkowiec, Marcin Małysza, Adam Bitka and Mirosław Głowacki
Materials 2023, 16(21), 6837; https://doi.org/10.3390/ma16216837 - 24 Oct 2023
Cited by 2 | Viewed by 1021
Abstract
The aim of this research was to develop a solution based on existing methods and tools that would allow the automatic classification of selected images of cast iron microstructures. As part of the work, solutions based on artificial intelligence were tested and modified. [...] Read more.
The aim of this research was to develop a solution based on existing methods and tools that would allow the automatic classification of selected images of cast iron microstructures. As part of the work, solutions based on artificial intelligence were tested and modified. Their task is to assign a specific class in the analyzed microstructure images. In the analyzed set, the examined samples appear in various zoom levels, photo sizes and colors. As is known, the components of the microstructure are different. In the examined photo, there does not have to be only one type of precipitate in each photo that indicates the correct microstructure of the same type of alloy, different shapes may appear in different amounts. This article also addresses the issue of data preparation. In order to isolate one type of structure element, the possibilities of using methods such as HOG (histogram of oriented gradients) and thresholding (the image was transformed into black objects on a white background) were checked. In order to avoid the slow preparation of training data, our solution was proposed to facilitate the labeling of data for training. The HOG algorithm combined with SVM and random forest were used for the classification process. In order to compare the effectiveness of the operation, the Faster R-CNN and Mask R-CNN algorithms were also used. The results obtained from the classifiers were compared to the microstructure assessment performed by experts. Full article
Show Figures

Figure 1

Figure 1
<p>Examples of images of cast iron microstructures used in the research.</p>
Full article ">Figure 2
<p>Examples of images of cast iron microstructures used in the research.</p>
Full article ">Figure 3
<p>Example of operation of a classical detector for the SVM classifier.</p>
Full article ">Figure 4
<p>Loss function graph for overall loss.</p>
Full article ">Figure 5
<p>Operation of the Faster R-CNN detector on an image containing mostly form IV objects.</p>
Full article ">Figure 6
<p>Loss function graph for overall loss, network retraining.</p>
Full article ">Figure 7
<p>Operation of the Mask R-CNN detector on an image containing mostly form VI objects.</p>
Full article ">
18 pages, 5568 KiB  
Article
A Method for Extracting Contours of Building Facade Hollowing Defects Using Polarization Thermal Images Based on Improved Canny Algorithm
by Darong Zhu, Jianguo Li, Fangbin Wang, Xue Gong, Wanlin Cong, Ping Wang and Yanli Liu
Buildings 2023, 13(10), 2563; https://doi.org/10.3390/buildings13102563 - 10 Oct 2023
Cited by 3 | Viewed by 1487
Abstract
During the service process of high-rise buildings, hollowing defects may be produced in the decorative layer, which not only affect the appearance, but also create a safety hazard of wall covering and shattered plaster peeling. Numerous studies have shown that hollowing can be [...] Read more.
During the service process of high-rise buildings, hollowing defects may be produced in the decorative layer, which not only affect the appearance, but also create a safety hazard of wall covering and shattered plaster peeling. Numerous studies have shown that hollowing can be detected using infrared thermal imagery under normal conditions. However, it is difficult to detect the edge and calculate the area of the hollowing on an exterior facade accurately because of the low contrast and fuzzy boundaries of the obtained infrared thermal images. To address these problems, a method for extracting the contours of building facade hollowing defects using polarization thermal images based on an improved Canny algorithm has been proposed in this paper. Firstly, the principle of thermal polarization imaging was introduced for hollowing detection. Secondly, considering the shortcomings of the Canny edge detection algorithm and the features of polarization thermal images, an improved Canny edge detection algorithm is proposed, including adaptive bilateral filtering to improve noise reduction ability while ensuring defect edges are not virtualized, Laplacian sharpening and histogram equalization to achieve contour sharpening and contrast enhancement, and eight-direction gradient templates for calculating image gradients, which make interpolation with non-maximum suppression more accurate, and the Tsallis entropy threshold segmentation algorithm based on the OTSU algorithm verification makes the image contour information more complete and accurate. Finally, a long-wave infrared polarization thermal imaging experimental platform was established and validation experiments were conducted. The experimental results demonstrate that the distinct, smooth, and precise location edges of the hollowing polarization infrared thermal images can be obtained, and the average error of the detected hollowing area is about 10% using the algorithm proposed in this paper. Full article
(This article belongs to the Section Construction Management, and Computers & Digitization)
Show Figures

Figure 1

Figure 1
<p>Schematic diagram of heat transfer process of experimental wall and the structural diagram of actual wall.</p>
Full article ">Figure 2
<p>Improved Canny algorithm operation flowchart.</p>
Full article ">Figure 3
<p>Improvement direction division diagram.</p>
Full article ">Figure 4
<p>Experimental wall: (<b>a</b>) Position distribution diagram of preset hollowing defects; (<b>b</b>) Size of preset hollowing defects; (<b>c</b>) Visual image; (<b>d</b>) IR image; (<b>e</b>) Polarization thermal image.</p>
Full article ">Figure 5
<p>Schematic diagram of experimental setup.</p>
Full article ">Figure 6
<p>Experimental equipment diagram. (1) A CCD long-wave infrared cooling camera; (2) A high-precision turntable; (3) An infrared metal grating polarizer.</p>
Full article ">Figure 7
<p>Comparison of image filtering processing effects: (<b>a</b>–<b>f</b>) The polarization thermal original and cropped images processed using Gaussian filtering; median filtering; mean filtering; bilateral filtering; and improved bilateral filtering.</p>
Full article ">Figure 8
<p>The polarization thermal image and image enhancement processing result.</p>
Full article ">Figure 9
<p>The schematic diagram of infrared thermal images and polarized thermal images.</p>
Full article ">Figure 10
<p>The results of hollowing defect images processed using the traditional Canny algorithm and morphological methods: (<b>a1</b>) Output edge of cropped infrared image; (<b>a2</b>) The result of (<b>a1</b>) after application of morphological methods; (<b>b1</b>) Output edge of cropped polarization image; (<b>b2</b>) The result of (<b>b1</b>) after application of morphological methods.</p>
Full article ">Figure 11
<p>Edge contour detection results of hollowing defects: (<b>a</b>–<b>f</b>) Roberts, Sobel, Prewitt, Log, Canny, and improved Canny algorithm.</p>
Full article ">Figure 12
<p>Morphological processing results: (<b>a1</b>–<b>f1</b>) Roberts, Sobel, Prewitt, Log, Canny, and improved Canny algorithm.</p>
Full article ">Figure 13
<p>The error rate of different algorithms in the size of hollowing defects.</p>
Full article ">
15 pages, 2479 KiB  
Article
A Robust Brain Tumor Detector Using BiLSTM and Mayfly Optimization and Multi-Level Thresholding
by Rabbia Mahum, Mohamed Sharaf, Haseeb Hassan, Lixin Liang and Bingding Huang
Biomedicines 2023, 11(6), 1715; https://doi.org/10.3390/biomedicines11061715 - 15 Jun 2023
Cited by 7 | Viewed by 1797
Abstract
A brain tumor refers to an abnormal growth of cells in the brain that can be either benign or malignant. Oncologists typically use various methods such as blood or visual tests to detect brain tumors, but these approaches can be time-consuming, require additional [...] Read more.
A brain tumor refers to an abnormal growth of cells in the brain that can be either benign or malignant. Oncologists typically use various methods such as blood or visual tests to detect brain tumors, but these approaches can be time-consuming, require additional human effort, and may not be effective in detecting small tumors. This work proposes an effective approach to brain tumor detection that combines segmentation and feature fusion. Segmentation is performed using the mayfly optimization algorithm with multilevel Kapur’s threshold technique to locate brain tumors in MRI scans. Key features are achieved from tumors employing Histogram of Oriented Gradients (HOG) and ResNet-V2, and a bidirectional long short-term memory (BiLSTM) network is used to classify tumors into three categories: pituitary, glioma, and meningioma. The suggested methodology is trained and tested on two datasets, Figshare and Harvard, achieving high accuracy, precision, recall, F1 score, and area under the curve (AUC). The results of a comparative analysis with existing DL and ML methods demonstrate that the proposed approach offers superior outcomes. This approach has the potential to improve brain tumor detection, particularly for small tumors, but further validation and testing are needed before clinical use. Full article
(This article belongs to the Special Issue Photodynamic Therapy in Cancer)
Show Figures

Figure 1

Figure 1
<p>The flow diagram for the proposed model.</p>
Full article ">Figure 2
<p>Some segmented samples from dataset. (<b>top</b>) Original images; (<b>bottom</b>) Segmented images.</p>
Full article ">Figure 3
<p>ResNet’s blocks.</p>
Full article ">Figure 4
<p>Some samples of brain MRI.</p>
Full article ">Figure 5
<p>Comparison with existing methods on the Harvard dataset [<a href="#B32-biomedicines-11-01715" class="html-bibr">32</a>,<a href="#B33-biomedicines-11-01715" class="html-bibr">33</a>,<a href="#B34-biomedicines-11-01715" class="html-bibr">34</a>,<a href="#B35-biomedicines-11-01715" class="html-bibr">35</a>,<a href="#B36-biomedicines-11-01715" class="html-bibr">36</a>,<a href="#B37-biomedicines-11-01715" class="html-bibr">37</a>].</p>
Full article ">Figure 6
<p>Comparison plot with existing techniques [<a href="#B18-biomedicines-11-01715" class="html-bibr">18</a>,<a href="#B39-biomedicines-11-01715" class="html-bibr">39</a>,<a href="#B40-biomedicines-11-01715" class="html-bibr">40</a>,<a href="#B41-biomedicines-11-01715" class="html-bibr">41</a>,<a href="#B42-biomedicines-11-01715" class="html-bibr">42</a>,<a href="#B43-biomedicines-11-01715" class="html-bibr">43</a>,<a href="#B44-biomedicines-11-01715" class="html-bibr">44</a>].</p>
Full article ">
14 pages, 4414 KiB  
Article
A Hybrid Cracked Tiers Detection System Based on Adaptive Correlation Features Selection and Deep Belief Neural Networks
by Ali Mohsin Al-juboori, Ali Hakem Alsaeedi, Riyadh Rahef Nuiaa, Zaid Abdi Alkareem Alyasseri, Nor Samsiah Sani, Suha Mohammed Hadi, Husam Jasim Mohammed, Bashaer Abbuod Musawi and Maifuza Mohd Amin
Symmetry 2023, 15(2), 358; https://doi.org/10.3390/sym15020358 - 29 Jan 2023
Cited by 9 | Viewed by 2073
Abstract
Tire defects are crucial for safe driving. Specialized experts or expensive tools such as stereo depth cameras and depth gages are usually used to investigate these defects. In image processing, feature extraction, reduction, and classification are presented as three challenging and symmetric ways [...] Read more.
Tire defects are crucial for safe driving. Specialized experts or expensive tools such as stereo depth cameras and depth gages are usually used to investigate these defects. In image processing, feature extraction, reduction, and classification are presented as three challenging and symmetric ways to affect the performance of machine learning models. This paper proposes a hybrid system for cracked tire detection based on the adaptive selection of correlation features and deep belief neural networks. The proposed system has three steps: feature extraction, selection, and classification. First, the oriented gradient histogram extracts features from the tire images. Second, the proposed adaptive correlation feature selection selects important features with a threshold value adapted to the nature of the images. The last step of the system is to predict the image category based on the deep belief neural networks technique. The proposed model is tested and evaluated using real images of cracked and normal tires. The experimental results show that the proposed solution performs better than the current studies in effectively classifying tire defect images. The proposed hybrid cracked tire detection system based on adaptive correlation feature selection and Deep Belief Neural Networks’ performance provided better classification accuracy (88.90%) than that of Belief Neural Networks (81.6%) and Convolution Neural Networks (85.59%). Full article
(This article belongs to the Section Computer)
Show Figures

Figure 1

Figure 1
<p>Proposed scenario of intelligent system for detection of cracked tires.</p>
Full article ">Figure 2
<p>Architecture of deep belief network.</p>
Full article ">Figure 3
<p>Graphical representation of restricted Boltzmann machines.</p>
Full article ">Figure 4
<p>Architecture of hybrid adaptive–correlation-based feature selection and deep belief neural networks.</p>
Full article ">Figure 5
<p>Graphical representation of histogram of oriented gradient features.</p>
Full article ">Figure 6
<p>Distribution value of the threshold (α) in the proposed adaptive correlation features selection.</p>
Full article ">Figure 7
<p>Flow of the proposed adaptive correlation-based and variance feature selection.</p>
Full article ">Figure 8
<p>Sample of images in the tires dataset.</p>
Full article ">Figure 9
<p>Comparison between naïve Bayes (NB), random forest (RF), and decision tree (DT), and deep belief neural network (DBN) in terms of true negative rate (TNR).</p>
Full article ">Figure 10
<p>Comparison between hybrid adaptive–correlation-based feature selection—deep belief neural network (H-DBN) and deep belief neural network (DBN) in terms of true positives rate (TPR).</p>
Full article ">Figure 11
<p>Comparison between hybrid adaptive–correlation-based feature selection—deep belief neural network (H-DBN) and deep belief neural network (DBN) in terms of receiver operating characteristic (ROC).</p>
Full article ">
16 pages, 25269 KiB  
Article
Improved Method Based on Retinex and Gabor for the Surface Defect Enhancement of Aluminum Strips
by Qi Zhang, Hongqun Tang, Yong Li, Bing Han and Jiadong Li
Metals 2023, 13(1), 118; https://doi.org/10.3390/met13010118 - 6 Jan 2023
Cited by 1 | Viewed by 1561
Abstract
Aiming at the problems of the blurred image defect contour and the surface texture of the aluminum strip suppressing defect feature extraction when collecting photos online in the air cushion furnace production line, we propose an algorithm for the surface defect enhancement and [...] Read more.
Aiming at the problems of the blurred image defect contour and the surface texture of the aluminum strip suppressing defect feature extraction when collecting photos online in the air cushion furnace production line, we propose an algorithm for the surface defect enhancement and detection of aluminum strips based on the Retinex theory and Gobar filter. The Retinex algorithm can enhance the information and detail part of the image, while the Gobar algorithm can maintain the integrity of the defect edges well. The method first improves the high-frequency information of the image using a multi-scale Retinex based on a Laplacian filter, scales the original image and the enhanced image, and enhances the contrast of the image by adaptive histogram equalization. Then, the image is denoised, and texture suppressed using median filtering and morphological operations. Finally, Gobar edge detection is performed on the obtained sample images by convolving the sinusoidal plane wave and the Gaussian kernel function in the null domain and performing double-threshold segmentation to extract and refine the edges. The algorithm in this paper is compared with histogram equalization and the Gaussian filter-based MSR algorithm, and the surface defects of aluminum strips are significantly enhanced for the background. The experimental results show that the information entropy of the aluminum strip material defect image is improved from 5.03 to 7.85 in the original image, the average gradient of the image is improved from 3.51 to 9.51 in the original image, the contrast between the foreground and background is improved from 16.66 to 117.53 in the original image, the peak signal-to-noise ratio index is improved to 24.50 dB, and the integrity of the edges is well maintained while denoising. This paper’s algorithm effectively enhances and detects the surface defects of aluminum strips, and the edges of defect contours are clearer and more complete. Full article
Show Figures

Figure 1

Figure 1
<p>Flow chart of the algorithm in this paper.</p>
Full article ">Figure 2
<p>Schematic diagram of the Retinex algorithm.</p>
Full article ">Figure 3
<p>Processing flow of the Retinex algorithm.</p>
Full article ">Figure 4
<p>Surface inspection plan layout.</p>
Full article ">Figure 5
<p>Improvement of the filter kernel.</p>
Full article ">Figure 6
<p>Schematic diagram of the choice of the scale and orientation of the Gabor filter.</p>
Full article ">Figure 7
<p>The experimental setup.</p>
Full article ">Figure 8
<p>Sample diagram of defects in aluminum strips: (<b>a,d</b>) blot; (<b>b,e</b>) lacerate; (<b>c,f</b>) scratch.</p>
Full article ">Figure 9
<p>Defect image histogram: (<b>a</b>) initial image; (<b>b</b>) histogram equalization; (<b>c</b>) MSR; (<b>d</b>) this article’s algorithm.</p>
Full article ">
26 pages, 9220 KiB  
Article
Detection of Risk Predictors of COVID-19 Mortality with Classifier Machine Learning Models Operated with Routine Laboratory Biomarkers
by Mehmet Tahir Huyut, Andrei Velichko and Maksim Belyaev
Appl. Sci. 2022, 12(23), 12180; https://doi.org/10.3390/app122312180 - 28 Nov 2022
Cited by 14 | Viewed by 1906
Abstract
Early evaluation of patients who require special care and who have high death-expectancy in COVID-19, and the effective determination of relevant biomarkers on large sample-groups are important to reduce mortality. This study aimed to reveal the routine blood-value predictors of COVID-19 mortality and [...] Read more.
Early evaluation of patients who require special care and who have high death-expectancy in COVID-19, and the effective determination of relevant biomarkers on large sample-groups are important to reduce mortality. This study aimed to reveal the routine blood-value predictors of COVID-19 mortality and to determine the lethal-risk levels of these predictors during the disease process. The dataset of the study consists of 38 routine blood-values of 2597 patients who died (n = 233) and those who recovered (n = 2364) from COVID-19 in August–December, 2021. In this study, the histogram-based gradient-boosting (HGB) model was the most successful machine-learning classifier in detecting living and deceased COVID-19 patients (with squared F1 metrics F12 = 1). The most efficient binary combinations with procalcitonin were obtained with D-dimer, ESR, D-Bil and ferritin. The HGB model operated with these feature pairs correctly detected almost all of the patients who survived and those who died (precision > 0.98, recall > 0.98, F12 > 0.98). Furthermore, in the HGB model operated with a single feature, the most efficient features were procalcitonin (F12 = 0.96) and ferritin (F12 = 0.91). In addition, according to the two-threshold approach, ferritin values between 376.2 μg/L and 396.0 μg/L (F12 = 0.91) and procalcitonin values between 0.2 μg/L and 5.2 μg/L (F12 = 0.95) were found to be fatal risk levels for COVID-19. Considering all the results, we suggest that many features combined with these features, especially procalcitonin and ferritin, operated with the HGB model, can be used to achieve very successful results in the classification of those who live, and those who die from COVID-19. Moreover, we strongly recommend that clinicians consider the critical levels we have found for procalcitonin and ferritin properties, to reduce the lethality of the COVID-19 disease. Full article
(This article belongs to the Special Issue Decision Support Systems for Disease Detection and Diagnosis)
Show Figures

Figure 1

Figure 1
<p>Pearson, Spearman and Kendall correlations of the SARS-CoV-2-RBV3 dataset for COVID-19 mortality-feature pairs.</p>
Full article ">Figure 2
<p>Spearman correlation analysis results for (<b>a</b>) the entire database, (<b>b</b>) survived COVID-19 class, and (<b>c</b>) non-survived COVID-19 class from the SARS-CoV-2-RBV3 dataset.</p>
Full article ">Figure 3
<p>Performance of ML models in classifying surviving and non-surviving COVID-19 patients, using the 34 features.</p>
Full article ">Figure 4
<p>F1 metrics for survived-COVID-19 class, calculated for original and SMOTE-balanced datasets.</p>
Full article ">Figure 5
<p>F1 metric for non-survived-COVID-19 class, calculated for original and SMOTE-balanced datasets.</p>
Full article ">Figure 6
<p>F1<sup>2</sup> metric of the HGB model according to each feature for the detection of surviving and non-surviving COVID-19 patients.</p>
Full article ">Figure 7
<p>The F1<sup>2</sup> metric for the classification of surviving and non-surviving COVID-19 patients, according to a single feature for the one-threshold approach, with dependency-type visualization (Type 1, Type 2).</p>
Full article ">Figure 8
<p>The F1<sup>2</sup> metric for the classification of surviving and non-surviving COVID-19 patients, according to a single feature for the two-threshold approach, with dependency-type visualization (Type 1, Type 2).</p>
Full article ">Figure 9
<p>Histogram distributions and F1<sup>2</sup> results of (<b>a</b>) procalcitonin, (<b>b</b>) ferritin and (<b>c</b>) fibrinogen properties, according to the single-cut-off value approach in estimating COVID-19 mortality. <span class="html-italic">V</span><sub>th</sub> (blue line) is the threshold for detecting COVID-19 mortality.</p>
Full article ">Figure 10
<p>Histogram distributions and F1<sup>2</sup> results of amylase feature according to two-threshold value approach in estimating COVID-19 mortality. <span class="html-italic">V</span><sub>th_1</sub> (pink line) and <span class="html-italic">V</span><sub>th_2</sub> (blue line) is the threshold for detecting COVID-19 mortality.</p>
Full article ">Figure 11
<p>F1<sup>2</sup> metric of SARS-CoV-2-RBV3 dataset for different models.</p>
Full article ">Figure 12
<p>Feature pairs with the highest F1<sup>2</sup> value that was found with the HGB classifier for detection of surviving and non-surviving COVID-19 patients.</p>
Full article ">Figure 13
<p>(<b>a</b>) Distribution of the procalcitonin feature in the original data of patients who survived and those who died from COVID-19, and the two-threshold value for this feature in classification. (<b>b</b>) The 1D masking technique for classifying patient-groups in the HGB model operated with the procalcitonin feature.</p>
Full article ">Figure 14
<p>Distributions of non-surviving and surviving COVID-19 patients over the original data on D-dimer-ferritin (<b>a</b>) and CK-MCH (<b>c</b>) feature pairs. The 2D-masking technique for patient-group classification of the HGB model operated with D-dimer-ferritin (<b>b</b>) and CK-MCH (<b>d</b>) feature pairs.</p>
Full article ">
29 pages, 3608 KiB  
Article
Monocular Camera Viewpoint-Invariant Vehicular Traffic Segmentation and Classification Utilizing Small Datasets
by Amr Yousef, Jeff Flora and Khan Iftekharuddin
Sensors 2022, 22(21), 8121; https://doi.org/10.3390/s22218121 - 24 Oct 2022
Cited by 2 | Viewed by 2422
Abstract
The work presented here develops a computer vision framework that is view angle independent for vehicle segmentation and classification from roadway traffic systems installed by the Virginia Department of Transportation (VDOT). An automated technique for extracting a region of interest is discussed to [...] Read more.
The work presented here develops a computer vision framework that is view angle independent for vehicle segmentation and classification from roadway traffic systems installed by the Virginia Department of Transportation (VDOT). An automated technique for extracting a region of interest is discussed to speed up the processing. The VDOT traffic videos are analyzed for vehicle segmentation using an improved robust low-rank matrix decomposition technique. It presents a new and effective thresholding method that improves segmentation accuracy and simultaneously speeds up the segmentation processing. Size and shape physical descriptors from morphological properties and textural features from the Histogram of Oriented Gradients (HOG) are extracted from the segmented traffic. Furthermore, a multi-class support vector machine classifier is employed to categorize different traffic vehicle types, including passenger cars, passenger trucks, motorcycles, buses, and small and large utility trucks. It handles multiple vehicle detections through an iterative k-means clustering over-segmentation process. The proposed algorithm reduced the processed data by an average of 40%. Compared to recent techniques, it showed an average improvement of 15% in segmentation accuracy, and it is 55% faster than the compared segmentation techniques on average. Moreover, a comparative analysis of 23 different deep learning architectures is presented. The resulting algorithm outperformed the compared deep learning algorithms for the quality of vehicle classification accuracy. Furthermore, the timing analysis showed that it could operate in real-time scenarios. Full article
(This article belongs to the Topic Applications in Image Analysis and Pattern Recognition)
Show Figures

Figure 1

Figure 1
<p>Proposed algorithm flow diagram.</p>
Full article ">Figure 2
<p>ROI extraction pipeline.</p>
Full article ">Figure 3
<p>A sample histogram of extracted sparse background for different dataset videos.</p>
Full article ">Figure 4
<p>Sample vehicles extracted from different videos in the dataset.</p>
Full article ">Figure 5
<p>Different output for ROI extraction pipeline stages for Video 1.</p>
Full article ">Figure 6
<p>Different output for ROI extraction pipeline stages for Video 4.</p>
Full article ">Figure 7
<p>Video 1 segmentation outputs comparison for different segmentation techniques.</p>
Full article ">Figure 8
<p>Video 2 segmentation outputs comparison for different segmentation techniques.</p>
Full article ">Figure 9
<p>Video 3 segmentation outputs comparison for different segmentation techniques.</p>
Full article ">Figure 10
<p>Video 4 segmentation outputs comparison for different segmentation techniques.</p>
Full article ">Figure 11
<p>ROC curves for Video 1 classification.</p>
Full article ">Figure 12
<p>ROC curves for Video 2 classification.</p>
Full article ">Figure 13
<p>ROC curves for Video 3 classification.</p>
Full article ">Figure 14
<p>ROC curves for Video 4 classification.</p>
Full article ">
18 pages, 4181 KiB  
Article
Development of an Optimal Algorithm for Detecting Damaged and Diseased Potato Tubers Moving along a Conveyor Belt Using Computer Vision Systems
by Sergey Alekseevich Korchagin, Sergey Timurovich Gataullin, Aleksey Viktorovich Osipov, Mikhail Viktorovich Smirnov, Stanislav Vadimovich Suvorov, Denis Vladimirovich Serdechnyi and Konstantin Vladimirovich Bublikov
Agronomy 2021, 11(10), 1980; https://doi.org/10.3390/agronomy11101980 - 30 Sep 2021
Cited by 22 | Viewed by 10537
Abstract
The article discusses the problem of detecting sick or mechanically damaged potatoes using machine learning methods. We proposed an algorithm and developed a system for the rapid detection of damaged tubers. The system can be installed on a conveyor belt in a vegetable [...] Read more.
The article discusses the problem of detecting sick or mechanically damaged potatoes using machine learning methods. We proposed an algorithm and developed a system for the rapid detection of damaged tubers. The system can be installed on a conveyor belt in a vegetable store, and it consists of a laptop computer and an action camera, synchronized with a flashlight system. The algorithm consists of two phases. The first phase uses the Viola-Jones algorithm, applied to the filtered action camera image, so it aims to detect separate potato tubers on the conveyor belt. The second phase is the application of a method that we choose based on video capturing conditions. To isolate potatoes infected with certain types of diseases (dry rot, for example), we use the Scale Invariant Feature Transform (SIFT)—Support Vector Machine (SVM) method. In case of inconsistent or weak lighting, the histogram of oriented gradients (HOG)—Bag-of-Visual-Words (BOVW)—neural network (BPNN) method is used. Otherwise, Otsu’s threshold binarization—a convolutional neural network (CNN) method is used. The first phase’s result depends on the conveyor’s speed, the density of tubers on the conveyor, and the accuracy of the video system. With the optimal setting, the result reaches 97%. The second phase’s outcome depends on the method and varies from 80% to 97%. When evaluating the performance of the system, it was found that it allows to detect and classify up to 100 tubers in one second, which significantly exceeds the performance of most similar systems. Full article
(This article belongs to the Section Innovative Cropping Systems)
Show Figures

Figure 1

Figure 1
<p>Potato Disease Identification (<b>1</b>. Late blight, <b>2</b>. Skin spot, <b>3</b>. Gangrene, <b>4</b>. Dry rot, <b>5</b>. Powdery scab, <b>6</b>. Tobacco Necrosis Virus, <b>7</b>. Common scab, <b>8</b>. Silver scurf, <b>9</b>. Potato Virus Y).</p>
Full article ">Figure 2
<p>Examples of potato tubers: (<b>1</b>–<b>3</b>) dry rotted tubers; (<b>4</b>–<b>6</b>) tubers affected by rodents; (<b>7</b>–<b>9</b>) healthy tubers.</p>
Full article ">Figure 3
<p>Original and processed images of potatoes (<b>1</b>. The original image; <b>2</b>. The image converted to grayscale; <b>3</b>. Threshold binarization applied (Otsu’s method); <b>4</b>. The adaptive binarization applied (Niblack’s method)).</p>
Full article ">Figure 4
<p>Inverted image and its processing (<b>1</b>. Inverted image; <b>2</b>. Inverted image converted to grayscale; <b>3</b>. Threshold binarization applied (Otsu’s method); <b>4</b>. Adaptive binarization applied (Niblack’s method)).</p>
Full article ">Figure 5
<p>Processing the image of potato tubers in <a href="#agronomy-11-01980-f002" class="html-fig">Figure 2</a> using the Sobel filter: (<b>1</b>–<b>3</b>) dry rotted tubers; (<b>4</b>–<b>6</b>) tubers affected by rodents; (<b>7</b>–<b>9</b>) healthy tubers.</p>
Full article ">Figure 6
<p>Processing the image of potato tubers in <a href="#agronomy-11-01980-f002" class="html-fig">Figure 2</a> by the HOG method: (<b>1</b>–<b>3</b>) dry rotted tubers; (<b>4</b>–<b>6</b>) tubers affected by rodents; (<b>7</b>–<b>9</b>) healthy tubers.</p>
Full article ">Figure 7
<p>Haar features.</p>
Full article ">Figure 8
<p>Identification of potato tubers by the Viola-Jones method.</p>
Full article ">Figure 9
<p>Identification of potato tubers by the Viola-Jones method: (<b>1</b>) damage localized in small areas, (<b>2</b>) damage in large areas.</p>
Full article ">Figure 10
<p>Steps to create Visual-Words.</p>
Full article ">Figure 11
<p>Generalized scheme of the algorithm.</p>
Full article ">
18 pages, 24345 KiB  
Article
Outdoor PV System Monitoring—Input Data Quality, Data Imputation and Filtering Approaches
by Sascha Lindig, Atse Louwen, David Moser and Marko Topic
Energies 2020, 13(19), 5099; https://doi.org/10.3390/en13195099 - 30 Sep 2020
Cited by 40 | Viewed by 5110
Abstract
Photovoltaic monitoring data are the primary source for studying photovoltaic plant behavior. In particular, performance loss and remaining-useful-lifetime calculations rely on trustful input data. Furthermore, a regular stream of high quality is the basis for pro-active operation and management activities which ensure a [...] Read more.
Photovoltaic monitoring data are the primary source for studying photovoltaic plant behavior. In particular, performance loss and remaining-useful-lifetime calculations rely on trustful input data. Furthermore, a regular stream of high quality is the basis for pro-active operation and management activities which ensure a smooth operation of PV plants. The raw data under investigation are electrical measurements and usually meteorological data such as in-plane irradiance and temperature. Usually, performance analyses follow a strict pattern of checking input data quality followed by the application of appropriate filter, choosing a key performance indicator and the application of certain methodologies to receive a final result. In this context, this paper focuses on four main objectives. We present common photovoltaics monitoring data quality issues, provide visual guidelines on how to detect and evaluate these, provide new data imputation approaches, and discuss common filtering approaches. Data imputation techniques for module temperature and irradiance data are discussed and compared to classical approaches. This work is intended to be a soft introduction into PV monitoring data analysis discussing best practices and issues an analyst might face. It was seen that if a sufficient amount of training data is available, multivariate adaptive regression splines yields good results for module temperature imputation while histogram-based gradient boosting regression outperforms classical approaches for in-plane irradiance transposition. Based on tested filtering procedures, it is believed that standards should be developed including relatively low irradiance thresholds together with strict power-irradiance pair filters. Full article
Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>General steps of monitoring data preparation.</p>
Full article ">Figure 2
<p>Original back of the module temperature measurements of mc-Si PV system; red—ambient temperature; blue—module temperature.</p>
Full article ">Figure 3
<p>Measured versus modeled (MARS regression model) module temperature of mc-Si PV system.</p>
Full article ">Figure 4
<p>Simplified modeling steps from global horizontal to in-plane irradiance.</p>
Full article ">Figure 5
<p>Data quality check figures for mc-Si PV system: (<b>a</b>) Normalized 1h energy values vs. time; (<b>b</b>) Normalized power-density plot—time of the day vs. day of the year; (<b>c</b>) Normalized 15 min power vs. in-plane irradiance; (<b>d</b>) Daily Performance Ratio vs. time.</p>
Full article ">Figure 6
<p>Exemplary data quality issues: (<b>a</b>) Imprecise irradiance sensor alignment (<span class="html-italic">P</span> vs. <math display="inline"><semantics> <msub> <mi>G</mi> <mrow> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">O</mi> <mi mathvariant="normal">A</mi> </mrow> </msub> </semantics></math>); (<b>b</b>) Inverter clipping (<span class="html-italic">E</span> vs. time); (<b>c</b>) Negative power values (<span class="html-italic">P</span> vs. <math display="inline"><semantics> <msub> <mi>G</mi> <mrow> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">O</mi> <mi mathvariant="normal">A</mi> </mrow> </msub> </semantics></math>); (<b>d</b>) Power data shift (<span class="html-italic">E</span> vs. time); (<b>e</b>) Data hole (<span class="html-italic">E</span> vs. time); (<b>f</b>) Inverter failure (daily <math display="inline"><semantics> <mrow> <mi>P</mi> <mi>R</mi> </mrow> </semantics></math> vs. time); (<b>g</b>) Irradiance sensor degradation (daily in-plane radiation vs. time); (<b>h</b>) Summer/Winter time shift and strong degradation (Normalized power-density plot).</p>
Full article ">Figure 7
<p><span class="html-italic">P</span> vs. <math display="inline"><semantics> <msub> <mi>G</mi> <mrow> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">O</mi> <mi mathvariant="normal">A</mi> </mrow> </msub> </semantics></math> and daily <math display="inline"><semantics> <mrow> <mi>P</mi> <mi>R</mi> </mrow> </semantics></math> vs. time plots for various filters applied to mc-Si PV system data—Filter 1: standard IEC 61724:2017 filter [<a href="#B10-energies-13-05099" class="html-bibr">10</a>,<a href="#B20-energies-13-05099" class="html-bibr">20</a>]; Filter 2: own filter used for performance loss analysis; Filter 3: clear-sky filter used in RdTools [<a href="#B43-energies-13-05099" class="html-bibr">43</a>].</p>
Full article ">
Back to TopTop