Search Results (7)

Advances in deep learning and computer vision have overcome many challenges inherent in the field of autonomous intelligent vehicles. To improve the detection accuracy and efficiency of EdgeBoard intelligent vehicles, we proposed an optimized design of EdgeBoard based on our PP-YOLOE+ model. This model innovatively introduces a composite backbone network, incorporating deep residual networks, feature pyramid networks, and RepResBlock structures to enrich environmental perception capabilities through the advanced analysis of sensor data. The incorporation of an efficient task-aligned head (ET-head) in the PP-YOLOE+ framework marks a pivotal innovation for precise interpretation of sensor information, addressing the interplay between classification and localization tasks with high effectiveness. Subsequent refinement of target regions by detection head units significantly sharpens the system’s ability to navigate and adapt to diverse driving scenarios. Our innovative hardware design, featuring a custom-designed mainboard and drive board, is specifically tailored to enhance the computational speed and data processing capabilities of intelligent vehicles. Furthermore, the optimization of our Pos-PID control algorithm allows the system to dynamically adjust to complex driving scenarios, significantly enhancing vehicle safety and reliability. Besides, our methodology leverages the latest technologies in edge computing and dynamic label assignment, enhancing intelligent vehicles’ operations through seamless sensor integration. Our custom dataset, specifically designed for this study, includes 4777 images captured by intelligent vehicles under a variety of environmental and lighting conditions. The dataset features diverse scenarios and objects pertinent to autonomous driving, such as pedestrian crossings and traffic signs, ensuring a comprehensive evaluation of the model’s performance. We conducted extensive testing of our model on this dataset to thoroughly assess sensor performance. Evaluated against metrics including accuracy, error rate, precision, recall, mean average precision (mAP), and F1-score, our findings reveal that the model achieves a remarkable accuracy rate of 99.113%, an mAP of 54.9%, and a real-time detection frame rate of 192 FPS, all within a compact parameter footprint of just 81 MB. These results demonstrate the superior capability of our PP-YOLOE+ model to integrate sensor data, achieving an optimal balance between detection accuracy and computational speed compared with existing algorithms. Full article

Achieving a balance between spectral resolution and spatial resolution in multi-spectral remote sensing images is challenging due to physical constraints. Consequently, pan-sharpening technology was developed to address this challenge. While significant progress was recently achieved in deep-learning-based pan-sharpening techniques, most existing deep learning approaches face two primary limitations: (1) convolutional neural networks (CNNs) struggle with long-range dependency issues, and (2) significant detail loss during deep network training. Moreover, despite these methods’ pan-sharpening capabilities, their generalization to full-sized raw images remains problematic due to scaling disparities, rendering them less practical. To tackle these issues, we introduce in this study a multi-spectral remote sensing image fusion network, termed TAMINet, which leverages a two-stream coordinate attention mechanism and multi-detail injection. Initially, a two-stream feature extractor augmented with the coordinate attention (CA) block is employed to derive modal-specific features from low-resolution multi-spectral (LRMS) images and panchromatic (PAN) images. This is followed by feature-domain fusion and pan-sharpening image reconstruction. Crucially, a multi-detail injection approach is incorporated during fusion and reconstruction, ensuring the reintroduction of details lost earlier in the process, which minimizes high-frequency detail loss. Finally, a novel hybrid loss function is proposed that incorporates spatial loss, spectral loss, and an additional loss component to enhance performance. The proposed methodology’s effectiveness was validated through experiments on WorldView-2 satellite images, IKONOS, and QuickBird, benchmarked against current state-of-the-art techniques. Experimental findings reveal that TAMINet significantly elevates the pan-sharpening performance for large-scale images, underscoring its potential to enhance multi-spectral remote sensing image quality. Full article

The grinding wheel topography influences the cutting performance and thus the economic efficiency of a grinding process. In contrary to conventional grinding wheels, super abrasive grinding wheels should undergo an additional sharpening process after the initial profiling process to obtain a suitable microstructure of the grinding wheel. Due to the lack of scientific knowledge, the sharpening process is mostly performed manually in industrial practice. A CNC-controlled sharpening process can not only improve the reproducibility of grinding processes but also decrease the secondary processing time and thereby increase the economic efficiency significantly. To optimize the sharpening process, experimental investigations were carried out to identify the significant sharpening parameters influencing the grinding wheel topography. The sharpening block width l_Sb, the grain size of the sharpening block d_kSb and the area-related material removal in sharpening V’’_Sb were identified as the most significant parameters. Additional experiments were performed to further quantify the influence of the significant sharpening parameters. Based on that, a process model was developed to predict the required sharpening parameters for certain target topographies. By using the process model, constant work results and improved process reliability can be obtained. Full article

Lung and colon cancers are two of the most common causes of death and morbidity in humans. One of the most important aspects of appropriate treatment is the histopathological diagnosis of such cancers. As a result, the main goal of this study is to use a multi-input capsule network and digital histopathology images to build an enhanced computerized diagnosis system for detecting squamous cell carcinomas and adenocarcinomas of the lungs, as well as adenocarcinomas of the colon. Two convolutional layer blocks are used in the proposed multi-input capsule network. The CLB (Convolutional Layers Block) employs traditional convolutional layers, whereas the SCLB (Separable Convolutional Layers Block) employs separable convolutional layers. The CLB block takes unprocessed histopathology images as input, whereas the SCLB block takes uniquely pre-processed histopathological images. The pre-processing method uses color balancing, gamma correction, image sharpening, and multi-scale fusion as the major processes because histopathology slide images are typically red blue. All three channels (Red, Green, and Blue) are adequately compensated during the color balancing phase. The dual-input technique aids the model’s ability to learn features more effectively. On the benchmark LC25000 dataset, the empirical analysis indicates a significant improvement in classification results. The proposed model provides cutting-edge performance in all classes, with 99.58% overall accuracy for lung and colon abnormalities based on histopathological images. Full article

The paper proposes a cost overrun risks prediction model, the structure of which is based on the fuzzy inference model of Mamdani. The model consists of numerous inputs and one output (MISO, multi-input-single-output), based on processes running consecutively in three blocks (the fuzzy block, the interference block, and the block of sharpening the representative output value). The input variables of the model include the share of element costs in the building costs (SE), predicted changes in the number of works (WC), and expected changes in the unit price (PC). The developed rule base makes it possible to determine the risk of cost overruns in the following categories: “very low”, “quite low”, “average”, “quite high”, and “very high”. Twenty-seven rules were assumed in the interference block. The operation of the model was illustrated by the example of selected elements of a road object and was validated by checking the correctness of the assumptions made at the design stage of the inference block rule base. It has been proven that with the increase of the share of element costs in the building costs (SE), predicted changes in the number of works (WC), and expected changes in the unit price (PC), the value of the risk exceeding the costs of a given element of the construction project (R) increases naturally and smoothly. It was emphasized in the conclusions that the cost overrun risks prediction model is intended for general contractors who subcontract many stages of works to their subcontractors in accordance with the agreed division into work elements. Full article

To assess the risk of project cost overrun, it is necessary to consider large amounts of symmetric and asymmetric data. This paper proposes a cost overrun risk prediction model, the structure of which is based on the fuzzy inference model of Mamdani. The model consists of numerous inputs and one output (multi-input-single-output (MISO)), based on processes running consecutively in three blocks (the fuzzy block, the interference block, and the block of sharpening the representative output value). The input variables of the model include the share of element costs in the building costs (SE), predicted changes in the number of works (WC), and expected changes in the unit price (PC). For the input variable SE, it is proposed to adjust the fuzzy set shapes to the type of building object. Single-family residential buildings, multi-family residential buildings, office buildings, highways, expressways, and sports fields were analyzed. The initial variable is the value of the risk of exceeding the costs of a given element of a construction investment project (R). In all, 27 rules were assumed in the interference block. Considering the possibility of applying sharpening methods in the cost overrun risk prediction model, the following defuzzification methods were investigated: the first of maxima, middle of maxima, and last of maxima method, the center of gravity method, and the bisector area method. Considering the advantages and disadvantages, the authors assumed that the correct and basic defuzzification method in the cost overrun risk prediction model was the center of gravity method. In order to check the correctness of the assumption made at the stage of designing the rule database, result diagrams were generated for the relationships between the variable (R) and the input variables of individual types of buildings. The results obtained confirm the correctness of the assumed assumptions and allow to consider the input variable (SE), adjusted individually to the model for each type of construction object, as crucial in the context of the impact on the output value of the output variable (R). Full article

High-resolution multi-spectral images are desired for applications in remote sensing. However, multi-spectral images can only be provided in low resolutions by optical remote sensing satellites. The technique of pan-sharpening wants to generate high-resolution multi-spectral (MS) images based on a panchromatic (PAN) image and the low-resolution counterpart. The conventional deep learning based pan-sharpening methods process the panchromatic and the low-resolution image in a feedforward manner where shallow layers fail to access useful information from deep layers. To make full use of the powerful deep features that have strong representation ability, we propose a two-path network with feedback connections, through which the deep features can be rerouted for refining the shallow features in a feedback manner. Specifically, we leverage the structure of a recurrent neural network to pass the feedback information. Besides, a power feature extraction block with multiple projection pairs is designed to handle the feedback information and to produce power deep features. Extensive experimental results show the effectiveness of our proposed method. Full article