Real-time Target Research Articles

Surface defect detection of wind turbine based on lightweight YOLOv5s model

The surface defect detection and maintenance of wind turbines affect the efficiency of wind power generation. However, traditional detection methods suffer from insufficient accuracy and large numbers of calculations. To solve this problem, we propose a surface defect detection model of wind turbines based on the lightweight YOLOv5s. Firstly, the YOLOv5s backbone network was replaced by the MobileNetv3 lightweight network for feature extraction to coordinate and balance the lightweight and accuracy relationship of the model. The neck network was realized as a weighted bi-directional feature pyramid network to enhance the model’s multi-scale adaptability and improve its fusion speed and efficiency. To adaptively adjust the feature weights and further improve the feature extraction ability, the convolution block attention module was used as the space and channel attention mechanism. Finally, the loss function was changed from border regression to the efficient intersection over union loss, which improved the accuracy of the bounding box regression. Compared with the original YOLOv5s model, the mean average precision was increased by 5.51 %, and the detection time was reduced by 10.79 frames per second, which shows that the improved model can complete the real-time target detection task on embedded devices with high accuracy and low energy consumption.

Real-time underwater target detection for AUV using side scan sonar images based on deep learning

To overcome the shortcomings of traditional methods for underwater target detection in sonar images, a novel target detection algorithm based on deep learning is proposed for the real-time detection for an autonomous underwater vehicle equipped with a side scan sonar (SSS). First, the preprocessing of raw SSS images and data augmentation methods are proposed to improve the quality and quantity of the data set. Second, to improve the detection efficiency, a method based on threshold segmentation and pixel importance value is used to quickly determine whether there are suspected targets in the SSS images. Then, an MA-YOLOv7 network based on YOLOv7 with multi-scale information fusion and an attention mechanism is proposed to detect targets in the screened images, and a target localization method is proposed to obtain the location (latitude and longitude) of the target. Finally, the algorithm is verified through simulation experiments and field tests. The results demonstrate that the proposed algorithm achieves state-of-the-art performance and can be applied to real underwater tasks with a detection recall of 0.836 and time consumption of 0.355 s per image.

Real-Time Target Tracking and Localization in Augmented Reality Scenes with Virtual Reality-Based Computer Vision Technology

Real-Time Target Tracking and Localization in Augmented Reality Scenes with Virtual Reality-Based Computer Vision Technology

Target recognition and tracking system based on UAV platform

Machine learning can extract the features of the target from the known image dataset and detect the target quickly by using the trained deep neural network. Due to UAVs’ fast flight speed and long detection distance, real-time and accurate target detection requirements are relatively high. For such scenarios,a ground object recognition and tracking method that integrates the YOLOv5 target detection algorithm and KCF target tracking algorithm are proposed. The combination of the two algorithms can meet the requirements of recognition speed, recognition rate, and tracking accuracy in the current scene. Two algorithms are run simultaneously in the recognition process, the detection frame with the highest confidence is selected as the final detection result, and the detected target is tracked by the KCF target tracking algorithm based on correlation filtering. The experimental results show that this method not only meets the requirements of recognition distance, recognition accuracy, and recognition speed of ground target but also can track a ground target in real-time, with strong robustness and real-time performance.

Performance analysis of the coherent FMCW photonic radar system under the influence of solar noise

The real-time high-resolution target detection in radar applications has increased the demand for photonic radar systems due to their higher bandwidth and faster processing capabilities over conventional microwave radar. Additionally, photonic-based radar technology can revolutionize the limited performance of conventional microwave radar caused by various atmospheric attenuations. This article presents a frequency-modulated continuous-wave photonic radar system using coherent detection. The performance of the proposed system is theoretically investigated under the effect of various real-time atmospheric weather attenuations and the influence of solar background noise, which is further verified on the simulation platform. This study is conducted under different atmospheric weather conditions, such as clear, haze, and fog, to highlight the practical limitation of free-space links in the presence of solar background noise. The detected signal at different target ranges with and without solar noise is measured and then compared in terms of signal-to-noise ratio. Furthermore, the impact of solar noise on the system performance when the Sun is under different sky conditions and zenithal angles is also analyzed. The results presented here provide insights into designing photonic radar systems for practical applications like autonomous vehicle radar systems, self-traffic control, and navigation.

Deep Learning Techniques for Vehicle Detection and Classification from Images/Videos: A Survey

Detecting and classifying vehicles as objects from images and videos is challenging in appearance-based representation, yet plays a significant role in the substantial real-time applications of Intelligent Transportation Systems (ITSs). The rapid development of Deep Learning (DL) has resulted in the computer-vision community demanding efficient, robust, and outstanding services to be built in various fields. This paper covers a wide range of vehicle detection and classification approaches and the application of these in estimating traffic density, real-time targets, toll management and other areas using DL architectures. Moreover, the paper also presents a detailed analysis of DL techniques, benchmark datasets, and preliminaries. A survey of some vital detection and classification applications, namely, vehicle detection and classification and performance, is conducted, with a detailed investigation of the challenges faced. The paper also addresses the promising technological advancements of the last few years.

Vehicle Acoustic and Seismic Synchronization Signal Classification Using Long-Term Features

Field vehicle type recognition plays an essential role in border protection tasks. Acoustic and seismic sensors can effectively collect the signal of field vehicle targets in real time. Most vehicle temporal signal classification algorithms are based on extracting and identifying handcrafted features. These algorithms focus on the signal’s frequency domain characteristics and despise the signal’s time domain characteristics. In order to extract appropriate features, this paper proposes a Long-Term Correlation Feature Network (LTCFN) to perform field vehicle acoustic and seismic signal classification. The model includes AlexNet-type feature extractor and an overall classifier implemented by a Long-Short Term Memory(LSTM) network. We presents an intra-frame network and fusion method for extracting feature vector from signals. Meanwhile, an inter-frame classifier is proposed firstly for analysing the time-correlation of the feature map and overall classification. The experiments illustrate that the LTCFN has excellent recognition performance and anti-noise ability. The classification accuracy of the LTCFN can be increased to 96%. This paper also provides a new idea for ground target classification through inter-frame feature measurement.

Safety Helmet Wearing Detection Based on Jetson Nano and Improved YOLOv5

Aiming to address the current shortcomings of the existing safety helmet wearing detection algorithms, including a slow reasoning speed, a large model size, and high hardware requirements, this study proposes an improved safety helmet wearing detection network named YOLOv5-SN, which is suitable for embedded deployment on Jetson Nano. First, the backbone of the YOLOv5 network is modified using the model lightweight method introduced by ShuffleNetV2. Next, the model size and number of parameters in the trained model are reduced to about one-tenth of those of the YOLOv5 network, and the reasoning speed is improved by 72 ms/f when tested on Jetson Nano. Then, the modified model is optimized using the quantification and layer fusion operations, further reducing the computing power and accelerating the reasoning speed. Finally, the YOLOv5-SN network is obtained by improving the YOLOv5 model, and the optimized model is deployed on Jetson Nano for testing. The average reasoning speed of the YOLOv5s-SN network reaches 32.2 ms/f, which is 84.7 ms/f faster compared to that of the YOLOv5s model. This demonstrates an obvious advantage of the proposed model in reasoning speed compared to the existing YOLOv5 models. Finally, the proposed model can perform real-time and effective target detection on the Jetson Nano embedded terminal.

Fixed-Point Landing Method for Unmanned Aerial Vehicles Using Multi-Sensor Pattern Detection

Conventional autonomous landing systems for unmanned aerial vehicles (UAVs) mainly rely on global positioning systems (GPS) or a single visual sensor, which identifies specific markers and uses them to guide the UAV landing. However, such methods have low positioning accuracies and weak range recognition abilities, which limit their applicability. This report proposes a fixed-point landing method combining light detection and ranging (LIDAR) technology with a vision camera; the developed approach can be used to locate the landing area and enable a completely autonomous landing, without being affected by distance, weather, or time. First, we establish the mathematical relationship between the color of the mark and the reflection intensity of the laser point cloud. According to this relationship, UAVs can confirm their position relative to the landing point starting from a greater distance without being disturbed by the environmental light intensity. Meanwhile, the method uses the camera to detect the corners of the landing target, even obtaining the relative position at a close distance. Finally, the LIDAR and camera data are combined to detect the corner points of the markers, determine the position of the UAV relative to the center of the landing target in real time, and guide the landing.

Improved Convolutional Neural Network YOLOv5 for Underwater Target Detection Based on Autonomous Underwater Helicopter

To detect a desired underwater target quickly and precisely, a real-time sonar-based target detection system mounted on an autonomous underwater helicopter (AUH) using an improved convolutional neural network (CNN) is proposed in this paper. YOLOv5 is introduced as the basic CNN network because of its strength, lightweight and fast speed. Due to the turbidity and weak illumination of an undesirable underwater environment, some attention mechanisms are added, and the structure of YOLOv5 is optimized to improve the performance of the detector for sonar images with a 1–3% increment of mAP which can be up to 80.2% with an average speed of 0.025 s (40 FPS) in the embedded device. It has been verified both in the school tank and outdoor open water that the whole detection system mounted on AUH performs well and meets the requirements of real time and light weight using limited hardware.

Research on the Method of Improving iGPS Dynamic Tracking Accuracy Based on Theoretical Trajectory Backward Compensation

As a result of the ongoing upgrading of the manufacturing industries, higher and higher requirements are being proposed for assembly precision. In particular, in the industries that involve assembly of large-sized products, it is hoped that assembly accuracy can still be guaranteed as the product size increases. The emergence of the indoor Global Positioning System (iGPS) will enable this demand to be met. In this work, we have conducted in-depth research on the measurement principles and the dynamic measurement performance of the iGPS measurement system. In addition, based on the motion equation for the tracked target, an analytical method to calculate the compensation amount required for dynamic measurements is proposed. This method is suitable for accurate measurement of the actual trajectory of a target in real time when the theoretical trajectory of this tracked target is known. We set up a simulation environment in MATLAB and tested the proposed method. The simulation testing proved that the method can compensate effectively for errors in iGPS dynamic measurement.

Task Scheduling for Distributed AUV Network Target Hunting and Searching: An Energy-Efficient AoI-Aware DMAPPO Approach

In this article, we aim to design a task scheduling scheme for the underwater multi-objective task of target hunting and environmental search. A distributed autonomous underwater vehicle (AUV) network is deployed to perform the task, where AUVs equipped with sensors can cooperatively search the environment and hunt the target by sharing local information. To achieve efficient exploration of the overall environment by the AUV network, we design an intra-network cooperative searching approach based on the age of information (AoI). Besides, it is critical to conceive an energy-efficient mechanism due to the energy constraints of AUVs and the difficulty of sustainable energy supply. To address the aforementioned issues, we propose an energy-efficient distributed multi-agent proximal policy optimization (DMAPPO) scheme to perform real-time AUV target hunting and environment searching in underwater turbulent fields. The proposed scheme can adjust the number of AUVs assigned to each objective according to practical requirement and residual energy. Distributed AUVs can make decisions autonomously and cooperatively complete the task efficiently through limited information interaction. In addition, we derive a lower bound on the policy improvement of MAPPO. Moreover, our simulation results demonstrate that the proposed scheme outperforms the standard algorithms in terms of hunting efficiency, degree of searching, and network energy efficiency.

PO-1901 Comparing LSTM networks for real-time target segmentation prediction on low-field MR-linacs

PO-1901 Comparing LSTM networks for real-time target segmentation prediction on low-field MR-linacs

High-precision real-time UAV target recognition based on improved YOLOv4

In recent years, unmanned aerial vehicles (UAVs) have gained widespread use in both military and civilian fields with the advancement of aviation technology and improved communication capabilities. However, the phenomenon of unauthorized UAV flights, or “black flying”, poses a serious threat to the safe flight of aircraft in airspace and public safety. To effectively interfere with and attack UAV targets, it is crucial to enhance the detection and identification of “low, slow and small” UAVs. This study focuses on achieving high-precision and lightweight detection and identification of four-rotor, six-rotor, and fixed-wing UAVs in low-altitude complex environments. By combining deep learning target detection with superresolution feature enhancement, a lightweight UAV detection model is designed and field-tested for verification. To address the challenge of detecting small UAV targets with limited information, the feature fusion network is enhanced based on the traditional YOLOv4 algorithm to improve the detection ability of small targets via small target enhancement and candidate box adjustment. The feasibility of the improved network is quantitatively and qualitatively analyzed. Channel pruning and layer pruning are then applied to the network, significantly reducing its depth and width and realizing a lightweight network. Finally, reasoning quantification is conducted on the embedded platform to enable end-side deployment of the target detection algorithm.

Improved Real Time Target Tracking System Based on Cam-shift and Kalman Filtering Techniques

This paper presents an improved real time target tracking system based on Cam-Shift and advanced Kalman filter algorithms. The paper, proposes a simple low implementation cost scheme while keeping the high tracking performance. Two degrees of freedom target tracking system based on two hybrid stepper motors with their drive systems is fully designed and implemented to improve the tracking accuracy while keeping low-cost. The potential of the proposed algorithm lies in its ability to accurately detect, track and produce firing commands on the designated target while utilizing low cost on the shelf embedded Microcontroller boards. A low cost Universal Serial Port (USB) camera is used and integrated with the proposed algorithm in order to detect and track maneuvering targets. A series of experiments for stationary and moving objects are carried out for testing the proposed system efficiency. The proposed system enables accurate performance for tracking static and moving objects. Hence it provides a simple and low implementation cost scheme for real time object tracking applications.

Correlation Filtering Algorithm of Infrared Spectral Data for Dim Target Tracking

The correlation filtering algorithm of infrared spectral data for dim and small target tracking is studied to improve the tracking accuracy of small and weak targets and to track small and weak targets in real time. After the image noise reduction processing by the mean shift filtering algorithm, the infrared small and weak target image data model is constructed by using the denoised infrared small and weak target image. And the brightness value and position of unknown small and weak targets are obtained. The tracking and measurement model of small and weak targets is built. The joint probabilistic data association algorithm is used to calculate the probability that each measurement is associated with its possible source targets, and the particle filter is used to update the tracking status of small and weak targets to achieve real-time tracking of small and weak targets. The experimental results show that the algorithm can enhance the edge contour information of small and weak images, so as to accurately track small and weak targets moving in any track, and has good real-time tracking and accuracy. There is a small deviation between the tracking track of weak and small targets tracked by the algorithm and the actual track, and the root mean square difference of tracking weak and small targets is within 2 pixels. In addition, the detection probability of detecting weak and small targets is less affected by SNR environmental factors.

A real time target face tracking algorithm based on saliency detection and Camshift

A real time target face tracking algorithm based on saliency detection and Camshift

Evaluating motion of pancreatic tumors and anatomical surrogates using cine MRI in 0.35T MRgRT under free breathing conditions.

Treatment tolerability is a significant limitation to pancreatic cancer treatment with radiotherapy due to proximity to highly radiosensitive organs and respiratory motion necessitating expanded target margins. Further, pancreatic tumors are difficult to visualize on conventional radiotherapy systems. Surrogates are often used to locate the tumor but are often inconsistent and do not provide strong positional relations throughout the respiratory cycle. This work utilizes a retrospective dataset of 45 pancreatic cancer patients treated on an MR-Linac system with cine MRI acquired for real-time target tracking. We investigated intra-fraction motion of tumors and two abdominal surrogates, leading to prediction models between the tumor and surrogate. Patient specific motion evaluation and prediction models were generated from 225 cine MRI series acquired during treatment. Tumor contours were used to evaluate the pancreatic tumor motion. Linear regression and principal component analysis (PCA) based models were used to predict tumor position from the anterior-posterior (AP) motion of the abdominal surface, the superior-inferior (SI) motion of the diaphragm, or a combination. Models were evaluated using mean squared error (MSE) and mean absolute error (MAE). Contour analysis showed the average pancreatic tumor motion range was 7.4±2.7mm and 14.9±5.8mm in the AP and SI directions, respectively. The PCA model had MSE of 1.4mm2 and 0.6mm2 , for the SI and AP directions, respectively, with both surrogates as inputs for the models. When only the abdomen surrogate was used, MSE was 1.3mm2 and 0.4mm2 in the SI and AP directions, while it was 0.4mm2 and 1.3mm2 when only the diaphragm surrogate was used. We evaluated intra-fraction pancreatic tumor motion and demonstrated prediction models between the tumor and surrogate. The models calculated the pancreatic tumor position from diaphragm, abdominal, or both contours within standard pancreatic cancer target margin, and the process could be applied to other disease sites in the abdominothoracic cavity.

Inference-Optimized High-Performance Photoelectric Target Detection Based on GPU Framework

Deep learning has better detection efficiency than typical methods in photoelectric target detection. However, classical CNNs on GPU frameworks consume too much computing power and memory resources. We propose a multi-stream inference-optimized TensorRT (MSIOT) method to solve this problem effectively. MSIOT uses knowledge distillation to effectively reduce the number of model parameters by layer guidance between CNNs and lightweight networks. Moreover, we use the TensorRT and multi-stream mode to reduce the number of model computations. MSIOT again increases inference speed by 9.3% based on the 4.3–7.2× acceleration of TensorRT. The experimental results show that the model’s mean average accuracy, precision, recall, and F1 score after distillation can reach up to 94.20%, 93.16%, 95.4%, and 94.27%, respectively. It is of great significance for designing a real-time photoelectric target detection system.

Abstract 2759: Real-time cellular target engagement and protein quantification for drug discovery

Abstract Preclinical assessment of direct drug-target engagement and analysis of binding kinetics within the cellular environment is essential for development of safer and more effective therapeutics. Infusion of such critical data into early drug discovery would significantly reduce the failure rate of new drug candidates, accelerate drug discovery, and validate repurposing of existing drugs. Described here is a novel drug-target engagement technology that can sensitively interrogate direct binding of drugs to cellular, bacterial, or viral protein targets within the physiological environment. Micro-Tag cell target engagement technology is based on complementation of a small 15-amino acid subunit with a large subunit into an active RNA-processing enzyme. The small subunit can be cloned to any drug target for transient or stable expression using existing tools such as CRISPR. Upon complementation, the active enzyme cleaves a FRET-based oligonucleotide substrate resulting in rapid generation of fluorescent signal that can be quantified in real time. The Micro-Tag technology enables in-cell quantitation of drug target levels using qPCR systems. This is the next-generation of target engagement technology that allows for real-time monitoring of drug-target interaction in the cell. We show here data demonstrating direct engagement of several reference compounds and novel small molecules with high-profile cancer targets such as K-RAS, MTH1, EGFR, and UBE2N. Selectivity of these drugs to the targets in the cell is further delineated using their mutant counterparts as well as inactive stereoisomer compounds. The Micro-Tag cell target engagement technology provides the power of cell target engagement to a large family of target proteins. It can be employed for high-throughput screens directed at initial on-target ranking or med-chem optimization of drug candidates. It is amenable for interrogation of drug candidates across various modalities: small molecules, peptides, antibodies and PROTACs. Importantly, this next-generation target engagement technology seamlessly integrates with qPCR systems, fluorescence microscopy, live-cell microscopy, FACS analysis, and offers multiplexing capability, thus allowing for further mechanistic insight. Citation Format: Ivan Babic, Nikolas Bryan, Claire Cunningham, Avery Sampson, Daniel Starczynowski, Elmar Nurmemmedov. Real-time cellular target engagement and protein quantification for drug discovery [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2759.