Average Frame Research Articles

Navigation path extraction for inter-row robots in Panax notoginseng shade house based on Im-YOLOv5s.

The accurate extraction of navigation paths is crucial for the automated navigation of agricultural robots. Navigation line extraction in complex environments such as Panax notoginseng shade house can be challenging due to factors including similar colors between the fork rows and soil, and the shadows cast by shade nets. In this paper, we propose a new method for navigation line extraction based on deep learning and least squares (DL-LS) algorithms. We improve the YOLOv5s algorithm by introducing MobileNetv3 and ECANet. The trained model detects the seven-fork roots in the effective area between rows and uses the root point substitution method to determine the coordinates of the localization base points of the seven-fork root points. The seven-fork column lines on both sides of the plant monopoly are fitted using the least squares method. The experimental results indicate that Im-YOLOv5s achieves higher detection performance than other detection models. Through these improvements, Im-YOLOv5s achieves a mAP (mean Average Precision) of 94.9%. Compared to YOLOv5s, Im-YOLOv5s improves the average accuracy and frame rate by 1.9% and 27.7%, respectively, and the weight size is reduced by 47.9%. The results also reveal the ability of DL-LS to accurately extract seven-fork row lines, with a maximum deviation of the navigation baseline row direction of 1.64°, meeting the requirements of robot navigation line extraction. The results shows that compared to existing models, this model is more effective in detecting the seven-fork roots in images, and the computational complexity of the model is smaller. Our proposed method provides a basis for the intelligent mechanization of Panax notoginseng planting.

New method for video stream brightness stabilization: algorithms and performance evaluation

Subject of study. In this paper, for the first time, an original method for estimating the change in the brightness of video data under the influence of changes in the lighting conditions of the scene and external noise is proposed. Algorithms for stabilizing the brightness of video data are also proposed. An objective assessment of the quality of video data pre-processed is given. The purpose of the research is to create a methodology for analyzing the variability of video data parameters under the influence of negative factors and to develop effective algorithms for stabilizing the parameters of the received video stream. The reliability of the method is tested using real video recordings pictured through various conditions. Objectives: To determine the most universal, resistant to external influences, and informative indicator necessary for an objective assessment of the quality of video data under various shooting conditions and scene lighting features; develop and programmatically implement algorithms for stabilizing video parameters based on modern programming tools. Research methods. Statistical analysis and pre-processing of video stream parameters as a random spatio-temporal process, algorithms for processing video data by digital filtering, and adaptive stabilization of video stream parameters. Research results. It has been proposed and experimentally proven that the optimal indicator of video stream quality is the average frame brightness (AFB). An algorithm for spatiotemporal processing of video data is proposed that generates a sequence of AFB values from the original video stream. The paper also proposes digital algorithms for filtering and stabilizing the brightness of a video stream and investigates the effectiveness of their application. Conclusions. The scientific novelty of the results obtained lies in a new method for analyzing and evaluating the parameters of video surveillance data and algorithms for filtering and stabilizing the brightness of the video stream. The performance of the proposed algorithms has been tested on real data. The algorithms are implemented in the Python software environment using the functions of the OpenCV library.

Multi-stage rotors assembly of turbine-based combined cycle engine based on augmented reality

To improve the assembly accuracy and efficiency of turbine-based combined cycle engines, we propose a multi-stage rotors assembly method based on augmented reality (AR) for turbine-based combined cycle engines. Combining the high fidelity interaction and edge computing capabilities of AR device, the real-time assembly guidance of multi-stage rotors of turbine-based combined cycle engine is realized. A new multi-stage rotors assembly mathematical model is proposed by introducing clearance errors to achieve prediction of coaxiality and perpendicularity in multi-stage rotors assembly. A point cloud registration method with equal-area constraints is proposed based on the requirement of AR virtual-real registration, combined with the triangulated graph. The experiment selected the multi-stage rotors of a simulated combination cycle engine for assembly. The experimental results showed that our AR system has an average delay of 11 ms and an average frame rate of 45fps. The accuracy of virtual-real registration is superior to current mainstream registration methods. The coaxiality prediction error and the perpendicularity prediction error for multi-stage rotors assembly is 0.5 μm and 0.4 μm, respectively. This system provides an efficient and accurate assembly plan for the rotors assembly of turbine-based combined cycle engines, meeting the actual assembly requirements on site.

Research on machine vision and deep learning based recognition of cotton seedling aphid infestation level.

Aphis gossypii Glover is a major insect pest in cotton production, which can cause yield reduction in severe cases. In this paper, we proposed the A. gossypii infestation monitoring method, which identifies the infestation level of A. gossypii at the cotton seedling stage, and can improve the efficiency of early warning and forecasting of A. gossypii, and achieve precise prevention and cure according to the predicted infestation level. We used smartphones to collect A. gossypii infestation images and compiled an infestation image data set. And then constructed, trained, and tested three different A. gossypii infestation recognition models based on Faster Region-based Convolutional Neural Network (R-CNN), You Only Look Once (YOLO)v5 and single-shot detector (SSD) models. The results showed that the YOLOv5 model had the highest mean average precision (mAP) value (95.7%) and frames per second (FPS) value (61.73) for the same conditions. In studying the influence of different image resolutions on the performance of the YOLOv5 model, we found that YOLOv5s performed better than YOLOv5x in terms of overall performance, with the best performance at an image resolution of 640×640 (mAP of 96.8%, FPS of 71.43). And the comparison with the latest YOLOv8s showed that the YOLOv5s performed better than the YOLOv8s. Finally, the trained model was deployed to the Android mobile, and the results showed that mobile-side detection was the best when the image resolution was 256×256, with an accuracy of 81.0% and FPS of 6.98. The real-time recognition system established in this study can provide technical support for infestation forecasting and precise prevention of A. gossypii.

FSO-Based Space-Air-Ground Integrated Vehicular Networks: Cooperative HARQ With Rate Adaptation

Space-air-ground integrated vehicular networks (SAGIVN) have been widely envisioned as a promising network architecture for 6 G to support the Internet of Vehicles (IoV). In SAGIVN, free-space optics (FSO)-based, high altitude platform (HAP) relay-assisted low Earth orbit (LEO) satellite systems have recently attracted research efforts worldwide. Critical challenges in designing and implementing FSO-based SAGIVN include atmospheric turbulence, weather conditions, and pointing misalignment. This paper offers a comprehensive cross-layer design framework of error-control protocols with rate adaptation for FSO burst transmissions in HAP-aided SAGIVN. Remarkably, we propose a design of link-layer cooperative incremental redundancy (IR) hybrid automatic repeat request (HARQ)-based sliding window mechanism. An analytical channel model for HAP-aided LEO satellite to emerging unmanned aerial vehicles (UAVs) FSO links is provided. At the same time, the analysis can also be applied to other kinds of vehicles. The queuing behavior induced by both the cooperative IR-HARQ protocol and the rate adaptation scheme is analyzed with a Markov model. Several performance metrics are analytically obtained, including the average round-trip frame delay, throughput, and energy efficiency. The results quantitatively demonstrate the impact of atmospheric turbulence and weather conditions on the system performance and support the optimal selection of system parameters. Additionally, the effectiveness of the proposed system is numerically confirmed. Monte Carlo simulations are also performed to validate the accuracy of theoretical derivations.

A deep-learning framework running on edge devices for handgun and knife detection from indoor video-surveillance cameras

The early detection of handguns and knives from surveillance videos is crucial to enhance people’s safety. Despite the increasing development of Deep Learning (DL) methods for general object detection, weapon detection from surveillance videos still presents open challenges. Among these, the most significant are: (i) the very small size of the weapons with respect to the camera field of view and (ii) the need of a real-time feedback, even when using low-cost edge devices for computation. Complex and recently-developed DL architectures could mitigate the former challenge but do not satisfy the latter one. To tackle such limitation, the proposed work addresses the weapon-detection task from an edge perspective. A double-step DL approach was developed and evaluated against other state-of-the-art methods on a custom indoor surveillance dataset. The approach is based on a first Convolutional Neural Network (CNN) for people detection which guides a second CNN to identify handguns and knives. To evaluate the performance in a real-world indoor environment, the approach was deployed on a NVIDIA Jetson Nano edge device which was connected to an IP camera. The system achieved near real-time performance without relying on expensive hardware. The results in terms of both COCO Average Precision (AP = 79.30) and Frames per Second (FPS = 5.10) on the low-power NVIDIA Jetson Nano pointed out the goodness of the proposed approach compared with the others, encouraging the spread of automated video surveillance systems affordable to everyone.

Deep learning for real-time detection of nasopharyngeal carcinoma during nasopharyngeal endoscopy

Nasopharyngeal carcinoma (NPC) is known for high curability during early stage of the disease, and early diagnosis relies on nasopharyngeal endoscopy and subsequent pathological biopsy. To enhance the early diagnosis rate by aiding physicians in the real-time identification of NPC and directing biopsy site selection during endoscopy, we assembled a dataset comprising 2,429 nasopharyngeal endoscopy video frames from 690 patients across three medical centers. With these data, we developed a deep learning-based NPC detection model using the you only look once (YOLO) network. Our model demonstrated high performance, with precision, recall, mean average precision, and F1-score values of 0.977, 0.943, 0.977, and 0.960, respectively, for internal test set and 0.825, 0.743, 0.814, and 0.780 for external test set at 0.5 intersection over union. Remarkably, our model demonstrated a high inference speed (52.9 FPS), surpassing the average frame rate (25.0 FPS) of endoscopy videos, thus making real-time detection in endoscopy feasible.

Smart warehouse track identification based on Res2Net-YOLACT+HSV

In recent years, the rapid development of the courier industry, in order to solve the problem of the accumulation of goods in courier warehouses, low efficiency of manual sorting and transportation, many courier companies have introduced intelligent unmanned warehouse logistics systems for unmanned transport vehicles or unmanned inspection The study proposes a recognition method based on the combination of Res2Net-YOLACT and HSV to achieve accurate recognition of the route trajectory of unmanned vehicles in the logistics warehouse environment. After the detection of the YOLACT network with improved backbone network, the area of the track is extracted by HSV, and the extracted area is used as the object of secondary processing, i.e., the extracted track is then subjected to centerline extraction to form the actual track. By comparing the recognition effect of this algorithm on the test set under different direction shapes of tracks, the correct rate of Res2Net-YOLACT in the same experimental environment is 97.37%, and the speed of detecting a single image is 30.26 ms, and then the centerline extraction algorithm is ported to the Res2Net-YOLACT network, and the speed of detecting a single image is 37.26 ms. Compared with the 93.28% accuracy of the original YOLACT network, the speed of single image detection is improved by 3.6%. In addition, the centerline extraction algorithm designed using this study is less computationally intensive and less code engineering than the prevailing algorithm, increasing the memory footprint by only 3.2%. To verify the performance of this centerline extraction algorithm, a comparison between this algorithm and the commonly available algorithms was performed, which showed a 1.31% improvement in correctness and an 8.73% improvement in the speed of detecting videos compared to the general algorithm, indicating that the centerline extraction algorithm processed by this study has higher accuracy and real-time performance without significantly consuming more memory. In addition, to test the practicality of the algorithm, the time used by the algorithm to detect the same video on the embedded device jetson nano was counted, and the average frame rate was calculated to be 28FPS and the maximum frame rate was 33FPS, which can be achieved in real transportation applications.

Detection of laryngeal carcinoma during endoscopy using artificial intelligence.

The objective of this study was to assess the performance and application of a self-developed deep learning (DL) algorithm for the real-time localization and classification of both vocal cord carcinoma and benign vocal cord lesions. The algorithm was trained and validated upon a dataset of videos and photos collected from our own department, as well as an open-access dataset named "Laryngoscope8". The algorithm correctly localizes and classifies vocal cord carcinoma on still images with a sensitivity between 71% and 78% and benign vocal cord lesions with a sensitivity between 70% and 82%. Furthermore, the best algorithm had an average frame per second rate of 63, thus making it suitable to use in an outpatient clinic setting for real-time detection of laryngeal pathology. We have demonstrated that our developed DL algorithm is able to localize and classify benign and malignant laryngeal pathology during endoscopy.

The change in propagation speed of natural shear waves by altering preload might differentiate patients with increased myocardial stiffness

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Research Foundation Flanders - FWO. Background Shear wave (SW) elastography (SWE) is a novel ultrasound-based technique for the assessment of myocardial stiffness. SWE tracks the propagation of SWs travelling through the myocardium, induced by e.g. mitral valve closure (MVC). The speed of SWs is related to myocardial stiffness. As myocardial stiffness is defined by a non-linear stress-strain relation, it is dependent on the operating stress levels and thus left ventricular (LV) loading. As a result, elevated SW speed values could be caused by both increased LV preload or structural remodelling of the myocardium. In this way, performing SWE measurements at different loading states could potentially provide additional information aiding with the interpretation of these measurements. Purpose to investigate if performing SWE before and after a loading change could provide more information on tissue properties of the myocardium. Methods Twelve patients [age: 76(41–82)] receiving intermittent hemodialysis (HD) and 12 age-matched healthy volunteers [age: 67(47–71); p = 0.19] were included. Patients were scanned before and 1 hour after the start of HD. Parasternal long-axis views were acquired using an experimental ultrasound scanner with an average frame rate of 1086±185 Hz. Patients were divided into 2 equal groups based on the grade of diastolic dysfunction (DD), determined using conventional echo parameters (group 1: DD grade 0–1 and group 2: DD grade 2−3). SWs after MVC were visualized on M-mode maps along the septum which were colour coded for tissue acceleration (Figure A). The speed was semi-automatically calculated by measuring the slope of the SW. Results The average ultra-filtrated volume after 1h of HD was 0.64±.27 L, which did not significantly differ between patient groups (p = 0.44). Patients in group 2 had significantly higher indexed LV wall mass compared to group 1 (128±39 g/m² vs. 72±24 g/m²; p = 0.02), indicative of LV structural remodelling. SW speed after MVC was significantly different between healthy volunteers and the patient groups at baseline (p<0.01) (Figure B). After 1h of HD, SW speed decreased in both groups, with the largest decrease in group 2 (Figure B). Furthermore, SW speed values of patient group 1 decreased to values similar to healthy volunteers, while patient group 2 still had significantly increased SW propagation speed. Conclusions Overall, myocardial stiffness, assessed using SW propagation speed, of patients with end-stage kidney disease was increased compared to healthy volunteers. This could result from increased LV preload or structural remodelling. Our data indicate that in addition to single measurements, assessing the change in SW speed by altering the loading state might aid in differentiating these causal mechanisms underlying increased myocardial stiffness. Therefore, adding a loading challenge, e.g. leg lifting, when doing SWE could provide additional information. Further research needs to confirm these initial findings.

Deep learning based object detection and surrounding environment description for visually impaired people

Object detection, one of the most significant contributions of computer vision and machine learning, plays an immense role in identifying and locating objects in an image or a video. We recognize distinct objects and precisely get their information through object detection, such as their size, shape, and location. This paper developed a low-cost assistive system of obstacle detection and the surrounding environment depiction to help blind people using deep learning techniques. TensorFlow object detection API and SSDLite MobileNetV2 have been used to create the proposed object detection model. The pre-trained SSDLite MobileNetV2 model is trained on the COCO dataset, with almost 328,000 images of 90 different objects. The gradient particle swarm optimization (PSO) technique has been used in this work to optimize the final layers and their corresponding hyperparameters of the MobileNetV2 model. Next, we used the Google text-to-speech module, PyAudio, playsound, and speech recognition to generate the audio feedback of the detected objects. A Raspberry Pi camera captures real-time video where real-time object detection is done frame by frame with Raspberry Pi 4B microcontroller. The proposed device is integrated into a head cap, which will help visually impaired people to detect obstacles in their path, as it is more efficient than a traditional white cane. Apart from this detection model, we trained a secondary computer vision model and named it the “ambiance mode.” In this mode, the last three convolutional layers of SSDLite MobileNetV2 are trained through transfer learning on a weather dataset. The dataset comprises around 500 images from four classes: cloudy, rainy, foggy, and sunrise. In this mode, the proposed system will narrate the surrounding scene elaborately, almost like a human describing a landscape or a beautiful sunset to a visually impaired person. The performance of the object detection and ambiance description modes are tested and evaluated in a desktop computer and Raspberry Pi embedded system. Detection accuracy and mean average precision, frame rate, confusion matrix, and ROC curve measure the model's accuracy on both setups. This low-cost proposed system is believed to help visually impaired people in their day-to-day life.

Hierarchical Detection of Gastrodia elata Based on Improved YOLOX

Identifying the grade of Gastrodia elata in the market has low efficiency and accuracy. To address this issue, an I-YOLOX object detection algorithm based on deep learning and computer vision is proposed in this paper. First, six types of Gastrodia elata images of different grades in the Gastrodia elata planting cooperative were collected for image enhancement and labeling as the model training dataset. Second, to improve feature information extraction, an ECA attention mechanism module was inserted between the backbone network CSPDarknet and the neck enhancement feature extraction network FPN in the YOLOX model. Then, the impact of the attention mechanism and application position on model improvement was investigated. Third, the 3 × 3 convolution in the neck enhancement feature extraction network FPN and the head network was replaced by depthwise separable convolution (DS Conv) to reduce the model size and computation amount. Finally, the EIoU loss function was used to predict boundary frame regression at the output prediction end to improve the convergence speed of the model. The experimental results indicated that compared with the original YOLOX model, the mean average precision of the improved I-YOLOX network model was increased by 4.86% (97.83%), the model computation was reduced by 5.422 M (reaching 3.518 M), the model size was reduced by 20.6 MB (reaching 13.7 MB), and the image frames detected per second increased by 3 (reaching 69). Compared with other target detection algorithms, the improved model outperformed Faster R-CNN, SSD-VGG, YOLOv3s, YOLOv4s, YOLOv5s, and YOLOv7 algorithms in terms of mean average precision, model size, computation amount, and frames per second. The lightweight model improved the detection accuracy and speed of different grades of Gastrodia elata and provided a theoretical basis for the development of online identification systems of different grades of Gastrodia elata in practical production.

Extra-Parotid Pleomorphic Adenoma and Low-Grade Salivary Malignancy in the Head and Neck Region.

Background and objective Pleomorphic adenoma is the most common major salivary gland neoplasm. Around 80% of them arise in the parotid glands, whereas 10% originate in the minor salivary glands. The pleomorphic adenoma of the extra-parotid site is defined by its location outside the primary salivary gland. The minor salivary gland adenomas occur at the hard and soft palate, lips, tongue, lacrimal glands, pharynx, larynx, paranasal sinus, and nasal cavity. Pleomorphic adenoma in parapharyngeal space may occur de novo or as an extension of the deep lobe of the parotid tumors. Our objective in this study wasto assess the location and presentations of extra-parotid pleomorphic adenomaand frequency of low-grade salivary gland malignancy diagnosed as pleomorphic adenomavia fine-needle aspiration cytology (FNAC)in the head and neck region andthe treatment outcomes after the resection of the tumors. Materials and methods This was a retrospective observational study. All patients with FNAC-diagnosedpleomorphic adenoma of extra-parotid locations of the head and neck region who underwent curative surgery in the Department of Otorhinolaryngology and Head and Neck Surgery at a rural tertiary care center between August 1997 and August 2022 were included in the study. Data on the symptoms, FNAC report, surgical techniques, pathological results, adjuvant therapy, andany recurrence were documented. Data were entered into a Microsoft Excelsheet and analyzed using IBM SPSS Statistics version22 software (IBM Corp., Armonk, NY). Results Our study included 23 patients, of which 14 were females and nine were males. The various sites of involvementwere as follows: parapharyngeal space (four), larynx (one), nasal septum (two), hard palate (five), soft palate (four), hard and soft palate (three), and submandibular salivary gland (four). Of note, 17.3% of the patients had local recurrence withan average time frame of three years post-surgery: 20% in patients with low-grade malignancy and 16.6% in patients with pleomorphic adenoma. Conclusion Extra-parotid pleomorphic adenomas are common and have a high malignant potential. The preferred choice of treatment for extra-parotid salivary tumors is complete resection with adequate clearance. Malignant pleomorphic adenomas may require stagingneck dissection and adjuvant treatment for a better prognosis.

Eye-Based Point Rendering for Dynamic Multiview Effects

Eye-based point rendering (EPR) can make multiview effects much more practical by adding eye (camera) buffer resolution efficiencies to improved view-independent rendering (iVIR). We demonstrate this very successfully by applying EPR to dynamic cube-mapped reflections, sometimes achieving nearly 7× speedups over iVIR and traditional multiview rendering (MVR), with nearly equivalent quality. Our application to omnidirectional soft shadows is less successful, demonstrating that EPR is most effective with larger shader loads and tight eye buffer to off-screen (render target) buffer mappings. This is due to EPR's eye buffer resolution constraints limiting points and shading calculations to the sampling rate of the eye's viewport. In a 2.48 million triangle scene with 50 reflective objects (using 300 off-screen views), EPR renders environment maps with a 49.40ms average frame time on an NVIDIA 1080 Ti GPU. In doing so, EPR generates up to 5x fewer points than iVIR, and regularly performs 50× fewer shading calculations than MVR.

The Analysis of Underwater Imagery System for Armor Unit Monitoring Application

The placement of armor units for breakwaters in Indonesia is still done manually, which depends on divers in each placement of the armor unit. The use of divers is less effective due to limited communication between divers and excavator operators, making divers in the water take a long time. This makes the diver's job risky and expensive. This research presents a vision system to reduce the diver's role in adjusting the position of each armor unit. This vision system is built with two cameras connected to a mini-computer. This system has an image improvement process by comparing three methods. The results obtained are an average frame per second is 20.71 without applying the method, 0.45 fps for using the multi-scale retinex with color restoration method, 16.75 fps for applying the Contrast Limited Adaptive Histogram Equalization method, 16.17 fps for applying the Histogram Equalization method. The image quality evaluation uses the underwater color quality evaluation with 48 data points. The method that has experienced the most improvement in image quality is multi-scale retinex with color restoration. Forty data have improved image quality with an average of 14,131, or 83.33%. The number of images that experienced the highest image quality improvement was using the multi-scale retinex with color restoration method. Meanwhile, for image quality analysis based on Underwater Image Quality Measures, out of a total of 48 images, the method with the highest value for image quality is the contrast limited adaptive histogram equalization method. 100% of images have the highest image matrix value with an average value is 33.014.

Design of Low Vision Electronic Glasses with Image Processing Capabilities Using Raspberry Pi

Poor vision is one of the most common eye health issues worldwide. Low vision patients are typically treated with optical devices or by substituting hearing or touch for visual capabilities. Head-mounted displays are currently the most promising form of low-vision assistive technology since they utilize the user's remaining natural visual capabilities. In this work, a prototype head-mounted display-based low-vision tool in the form of electronic glasses was designed utilizing a Raspberry Pi computer. The prototype was created using a Raspberry Pi 4 B coupled with cameras to allow real-time video acquisition. The LCD on the electronic eyewear frame as the camera showed the video recording. The prototype also included software utilizing five image processing modes—magnification, brightness enhancement, adaptive contrast enhancement, edge enhancement, and text detection and recognition- to help persons with limited vision acquire visual information more effectively. OpenCV was used with Python to create the software system. Average framerate measurements of 30–40 FPS for brightness and contrast improvement modes, 20 FPS for zooming and edge enhancement modes, and 1.3 FPS for text identification modes showed that the concept of electronic spectacles was successfully implemented in this research.

Comparison of burst properties between FRB 20190520B and FRB 20121102A

ABSTRACT A comparative analysis of the individual bursts between FRB 20190520B and FRB 20121102A is presented by compiling a sample of bursts in multiple wavelengths. It is found that the peak frequency (νp) distribution of the bursts of FRB 20190520B illustrates four discrete peaks in ∼1–6 GHz and their spectral width distribution can be fitted with a lognormal function peaking at 0.35 GHz. The discrete νp distribution and the narrow-banded spectral feature are analogous to FRB 20121102A. The burst duration of FRB 20190520B in the rest frame averages to 10.72 ms, longer than that of FRB 20121102A by a factor 3. The specific energy $\left(E_{\rm\mu_{\rm c}}\right)$ at 1.25 GHz of FRB 20190520B observed with the Five-hundred-meter Aperture Spherical Radio Telescope narrowly ranges in [0.4, 1] × 1038 erg, different from the bimodal $E_{\rm \mu _{\rm c}}$ distribution of FRB 20121102A. Assuming a Gaussian spectral profile of the bursts, our Monte Carlo simulation analysis suggests that a power law (PL) or a cut-off power-law (CPL) energy function can comparably reproduce the $E_{\rm \mu _{\rm c}}$ distribution of FRB 20190520B. The derived energy function index of the PL model is 4.46 ± 0.17, much steeper than that of FRB 20121102A ($1.82^{+0.10}_{-0.30}$ ). For the CPL model, we obtain an index of 0.47 and a cut-off energy of 7.4 × 1037 erg. Regarding the predicted νp distribution in 1–2 GHz, the CPL model is more preferred than the PL model. These results indicate that FRB 20190520B and FRB 20121102A share similar spectral properties, but their energy functions are intrinsically different.

A Dual-Input Moving Object Detection Method in Remote Sensing Image Sequences via Temporal Semantics

Moving object detection in remote sensing image sequences has been widely used in military and civilian fields. However, the complex background of remote sensing images and the small sizes of moving objects bring great difficulties for effective detection. To solve this problem, we propose a real-time moving object detection method for remote sensing image sequences. This method works by fusing the semantic information from a single image extracted by the object detection branch with the motion information of multiple frames extracted by the motion detection branch. Specifically, in the motion detection branch, we design a motion feature enhancement module (MFE) to improve the interframe motion information. Then, we design a Motion Information Extraction network (MIE) to extract motion information. Finally, the moving object information is directly output by fusing the motion and semantic information extracted by the object detection branch. Based on the experimental results of the two datasets, the proposed method achieves an accuracy rate of 93.21%, a recall rate of 92.72%, an average frame rate of 25.25 frames (fps), and a performance of 96.71% in terms of AP@0.5. The performance of the proposed method is better than that of other methods, and the overall detection effect is better; therefore, it meets the needs of the detection task.

Single shot multibox detector object detection based on attention mechanism and feature fusion

The single shot multibox detector (SSD) is one of the most important algorithms in single-stage target detection, having the characteristics of multiscale detection and rapid detection speed. However, the effective SSD feature layers are independent of one another, which can lead to object detection difficulties. To address this problem, we proposed an improved SSD object detection algorithm. First, the global attention mechanism (GAM)—which can enhance spatial and channel information—was introduced into the multiscale feature layer. The channel attention module of the GAM was improved. Second, a feature fusion module was introduced to strengthen the relationship between feature layers. Finally, the cross stage partial structure was introduced into the feature fusion module, and used to improve the model’s learning ability. For model training and detection based on the PASCAL VOC dataset, the mean average precision and frames per second obtained by the improved SSD algorithm were 84.67% and 18.67, respectively, which could effectively detect difficult targets.

A lightweight network of near cotton‐coloured impurity detection method in raw cotton based on weighted feature fusion

AbstractThe objective is to improve the recognition rate of white and near cotton‐coloured impurities in raw cotton against a single white visible light source background. A lightweight detection network model without anchor boxes based on improved YOLO v4‐tiny is proposed in this paper based on weighted feature fusion (WFF). The WFF strategy was used to improve the detection accuracy of the improved YOLO v4‐tiny algorithm. Meanwhile, to address the disadvantage that the anchor boxes obtained by the K‐means algorithm clustering do not have global features, the anchor‐free and tiny decoupled head schemes are used instead of the traditional coupled detection head. The improved algorithm was validated on the PASCAL VOC2012 dataset and the self‐built raw cotton impurity dataset collected by high‐speed camera. On the raw cotton impurity test set, compared to the original YOLO v4‐tiny model, the mean average precision (mAP) and frames per second (FPS) of the improved model increased by 10.35% and 6.9%, respectively. The proposed model detects white and near cotton‐coloured impurities with an accuracy of up to 98.78% and 98.00%. The experimental results show that the proposed method can effectively detect and identify impurities in raw cotton, which is of great practical significance for foreign matter detection and identification in cotton.