Traditional Target Detection Research Articles

Dim Space Target Detection via Convolutional Neural Network in Single Optical Image

Real-time dim space target detection is a significant challenge in space situation awareness. This paper proposes a single-frame space object segmentation and detection method based on deep learning. Firstly, Channel and Space Attention U-net (CSAU-Net) is presented based on space image features. We remove unnecessary feature layers and add attention modules in the traditional encoder and decoder structure to enhance feature fusion and better use original feature layers. The proposed network structure can achieve accurate segmentation of space objects with fewer data training. At the same time, we construct a space target dataset for training, which contains targets with different signal-to-noise ratios to enhance the generalization of convolutional neural networks. After obtaining the segmentation masks, a simple connected component labeling method is applied to extract the centroid of the space target. Experiments show that our approach can achieve an ideal segmentation effect when the signal-to-noise ratio of the space target in the simulated dataset is 0.3. In addition, the proposed algorithm can realize fast segmentation and achieve an accuracy of 98.5%, which is similar to the traditional multi-frame space target detection method in real space image detection.

Multifractal Correlation Analysis of Autoregressive Spectrum-Based Feature Learning for Target Detection Within Sea Clutter

Fractal theory has improved the target detection performance under sea clutter background. However, the traditional fractal methods in time domain or Fourier domain cannot accurately characterize the complex sea clutter properties, which leads to the degradation of the target detection performance under low signal-clutter-ratio (SCR) conditions. This article investigates the multifractal correlation property of sea clutter in the autoregressive (AR) spectrum domain for detection performance improvement. In this work, the traditional target detection problem is converted to a binary classification problem of sea clutter and targets. The refined fractal characteristics in various singularity scale interval and range bins are analyzed, and the AR singularity intensity correlation function width together with the accumulation area of AR multifractal correlation spectrum is extracted as intrinsic features. Then, a simple and efficient fully connected network is developed to realize the classification. Experimental results on real measured marine radar datasets demonstrate that the proposed method can increase the detection probability by about 15% than the state-of-art fractal-based methods under a low SCR condition.

License Plate Localization in Complex Environments Based on Improved GrabCut Algorithm

Aiming at the problem that the existed license plate detection method lacking of accuracy and speed, an improved lightweight detection algorithm for license plate detection in natural scenarios was proposed. First, the traditional GrabCut algorithm needs to interactively provide a candidate frame in order to perform the target detection work. We replace the candidate frame by introducing the Aspect ratio of the license plate as the foreground extraction feature to automate the detection of the license plate by GrabCut algorithm. Then, in order to improve the detection precision of traditional target detection algorithms, we introduced the Wiener filter, which is widely used in the field of digital signal processing, and Combine with Bernsen algorithm to complete image noise reduction. Finally, the algorithm was tested with the CCPD dataset, which contains many vehicle images from different complex natural scenes, especially low-resolution images. The experimental results shows that improved GrabCut algorithm achieves an average accuracy of 99.34% for license plate localization and a detection speed of 0.29s/frame, which has better accuracy and real-time performance compared with traditional GrabCut and other license plate localization algorithms.

Target Detection Based on Deep Learning

As an important research direction of computer vision, target detection has been widely used in face recognition, intelligent driving, robot navigation and other fields. In recent years, with the deepening research on deep learning, great progress has been made in the field of computer vision, such as image acquisition, image processing and target detection. Compared with the traditional target detection algorithm based on candidate regions, it has the problems of poor timeliness and slow detection speed. Recently, the popular target detection algorithm based on regression realizes the real sense of end-to-end detection and greatly improves the detection efficiency. However, the accuracy of small target detection and dense target detection has not been solved. In the future, we still need to improve the efficiency and accuracy of recognition on the existing basis, and solve the problem of small target and dense target detection to make it more widely used in practical application scenarios. In this paper, the principle, advantages and disadvantages, accuracy and other aspects of the above algorithms are introduced in detail, the problems existing in the target detection algorithm are summarized, and the future development direction has been prospected. In short, both algorithms have advantages and disadvantages, but the regression-based target detection algorithm has better practicality and development prospects.

LF-CNN: Deep Learning-Guided Small Sample Target Detection for Remote Sensing Classification

Target detection of small samples with a complex background is always difficult in the classification of remote sensing images. We propose a new small sample target detection method combining local features and a convolutional neural network (LF-CNN) with the aim of detecting small numbers of unevenly distributed ground object targets in remote sensing images. The k-nearest neighbor method is used to construct the local neighborhood of each point and the local neighborhoods of the features are extracted one by one from the convolution layer. All the local features are aggregated by maximum pooling to obtain global feature representation. The classification probability of each category is then calculated and classified using the scaled expected linear units function and the full connection layer. The experimental results show that the proposed LF-CNN method has a high accuracy of target detection and classification for hyperspectral imager remote sensing data under the condition of small samples. Despite drawbacks in both time and complexity, the proposed LF-CNN method can more effectively integrate the local features of ground object samples and improve the accuracy of target identification and detection in small samples of remote sensing images than traditional target detection methods.

Multiscale and Direction Target Detecting in Remote Sensing Images via Modified YOLO-v4

Traditional target detection algorithms have difficulty to adapt complex environmental changes and have limited applicable scenarios. However, the deep-learning-based target detection model can automatically learn with strong generalization capability. In this article, we choose a single-stage deep-learning-based target detection model for research based on the model’s real-time processing requirements and to improve the accuracy and the robustness of target detection in remote sensing images. In addition, we improve the YOLOv4 network and present a new approach. First, we propose a classification setting of the nonmaximum suppression threshold to increase the accuracy without affecting the speed. Second, we study the anchor frame allocation problem in YOLOv4 and propose two allocation schemes. The proposed anchor frame scheme also improves the detection performance, and experimental results on the DOTA dataset validate their effectiveness.

An End-to-End Anti-Jamming Target Detection Method Based on CNN

Target detection in the presence of main-lobe deception jamming has received intensive attention. Generally, traditional target detection with additive jamming can be regarded as an adjacent process of jamming suppression and target detection. Since target information can be lost in jamming suppression, the detection performance can degrade severely. Aiming at this issue, an end-to-end anti-jamming target detector is proposed based on convolutional neural network (CNN) in the presence of interrupted-sampling repeater jamming. The CNN method first converts the single-pulse compression filter into a series of feature extraction filters, accumulating the energy of various features in various dimensions and on multiple scales. Then, a feature fusion filter added after the feature extraction stage is utilized to suppress jamming features while simultaneously retaining the actual target features. The two stages can be achieved in the CNN structure, and all the filters in this method are obtained by training the CNN. Numerical simulations show the efficiency of the proposed method.

Multi-Dimensional Automatic Detection of Scanning Radar Images of Marine Targets Based on Radar PPInet

Traditional radar target detection algorithms are mostly based on statistical theory. They have weak generalization capabilities for complex sea clutter environments and diverse target characteristics, and their detection performance would be significantly reduced. In this paper, the range-azimuth-frame information obtained by scanning radar is converted into plain position indicator (PPI) images, and a novel Radar-PPInet is proposed and used for marine target detection. The model includes CSPDarknet53, SPP, PANet, power non-maximum suppression (P-NMS), and multi-frame fusion section. The prediction frame coordinates, target category, and corresponding confidence are directly given through the feature extraction network. The network structure strengthens the receptive field and attention distribution structure, and further improves the efficiency of network training. P-NMS can effectively improve the problem of missed detection of multi-targets. Moreover, the false alarms caused by strong sea clutter are reduced by the multi-frame fusion, which is also a benefit for weak target detection. The verification using the X-band navigation radar PPI image dataset shows that compared with the traditional cell-average constant false alarm rate detector (CA-CFAR) and the two-stage Faster R-CNN algorithm, the proposed method significantly improved the detection probability by 15% and 10% under certain false alarm probability conditions, which is more suitable for various environment and target characteristics. Moreover, the computational burden is discussed showing that the Radar-PPInet detection model is significantly lower than the Faster R-CNN in terms of parameters and calculations.

Underwater Target Detection Based on Deep Neural Network and Image Enhancement

Underwater target detection tasks refer to the detection of targets contained in underwater images. Unlike traditional target detection tasks, for underwater targets, due to factors such as illumination, camera shake, complex background interference, and diversification of target types, the effect of target detection will be affected. In this paper, we propose a target detection algorithm based on image enhancement and deep network. The algorithm first enhances the image data to obtain a better contrast, and then uses a deep learning algorithm to separate the target and the background to improve the detection performance of the target. Experimental results show that the algorithm can achieve better detection performance.

Hyperspectral target detection based on transform domain adaptive constrained energy minimization

Traditional hyperspectral target detection methods use spectral domain information for target recognition. Although it can effectively retain intrinsic characteristics of substances, targets in homogeneous regions still cannot be effectively recognized. By projecting the spectral domain features on the transform domain to increase the separability of background and target, fractional domain-based revised constrained energy minimization detector is proposed. Firstly, the fractional Fourier transform is adopted to project the original spectral information into the fractional domain for improving the separability of background and target. Then, a newly revised constrained energy minimization detector is performed, where sliding double window strategy is used to make the best of the local spatial statistical characteristics of testing pixel. In order to make the best of inner window information, the mean value of Pearson correlation coefficient is measured between prior target pixel and testing pixel associated with its four neighborhood pixels. Extensive experiments for four real hyperspectral scenes indicate that the performance of the proposed algorithm is excellent when compared with other related detectors.

Adaptive parameters optimization model with 3D information extraction for infrared small target detection based on particle swarm optimization algorithm

Traditional small target detection methods could not adjust the parameters adaptively to achieve small target detection in real scene applications. To solve this problem, a novel small target detection method named Adaptive parameters optimization model with three-dimensional (3D) Information Extraction (A3DIE) is proposed. In this method, a multi-objective particle swarm optimization algorithm is designed to optimize the parameters in the 3D Information Extraction (3DIE) method and realize the adaptability of detection method in different detection scenarios. In the optimization algorithm, an adaptive environment selection strategy is developed to enhance the evolutionary ability and obtain high-quality solution sets. Furthermore, a knee point selection strategy is designed to obtain the optimal small target detection method parameters. The experimental results indicate that our method can accurately and stably detect small targets in different real scenes compared with the baseline methods.

Random forest-based human pose detection system for through-the-wall radar

Nowadays, the detection, identification and classification of targets behind buildings in the process of wall penetration sensing is one of the main solution directions for detection activities. In this paper, a random forest-based human pose detection system for through-wall radar is proposed, aiming at optimizing the traditional through-wall radar target detection and identification by machine learning methods. The actual data acquisition is performed by UWB-MIMO through-wall radar system to construct multidimensional data and identify the pose. The experimental results show that the random forest method has high recognition performance by identifying multiple poses and has a pose resolution that traditional target recognition does not have.

Signal structure information-based target detection with a fully convolutional network

For target echoes, some structure information can be introduced by the signal processing techniques, such as matched filtering and coherent integration, and are usually omitted in traditional target detection (TTD) methods. The detection performance is supposed to be improved with consideration of these information. To deal with the randomness in the signal structure information (SSI) induced by the sampling and uncertain distribution of scatters, we resort to the data-driven method and propose a novel detection scheme. To make use of the SSI, a fully convolutional network (FCN) is designed to hierarchically learn the SSI. Simulation results show that the better detection performance can be obtained with the proposed SSI-based target detection method comparing to the TTD method. The justifications of using the SSI and the FCN are respectively investigated by considering the oversampling strategy and a post hoc visual explanation technique. Besides, the computational complexity is partially analyzed both in theory and in experiment.

Research on the Application of Computer CNN in Image Recognition

With the development of AI, deep learning (hereinafter referred to as DL) has been applied to many aspects, especially the application of CNN in image recognition (hereinafter referred to as IR). CNN is an important branch of DL, which has been deeply studied in IR. Through CNN, we can choose the right image in the image set, which can also be classified. With the different purposes of IR, the academia has developed a variety of different CNN models, which will better promote the application in IR. By increasing the training data, we can improve the recognition rate and accuracy of CNN, which will improve the efficiency of IR. IR based on CNN has broken the bottleneck of traditional target detection (hereinafter referred to as TD), which has become a mainstream algorithm in the field of TD. Firstly, this paper analyzes the basic theory and related models of CNN. Finally, this paper analyzes the application of CNN in IR.

Visual inspection for transformer insulation defects by a patrol robot fish based on deep learning

In order to effectively detect the internal insulation defects of large transformers, a miniature patrol robot fish is designed and used to observe the pressboard inside the transformer, which can visually inspect the insulation condition of the pressboard. In the process of visual inspection by a patrol robot fish, insulation defects observed in the transformer (such as discharge carbon marks, pressboard cracks, etc.) are small size, indistinctive colour contrast and different shapes. The key purpose of the patrol robot fish is to identify the defect targets and defect types intelligently and quickly from the images taken by the camera on the fish. Considering that there are abundant insulation defects in transformers and a lack of samples for learning, a vision detection method based on deep learning network is proposed in this study. The proposed method integrates the variable autoencoder into the traditional Faster-RCNN target detection network and constructs an improved Faster-RCNN that enhances feature extraction. This method expands the small-scale training sample set and improves the generalisation performance of the model. In order to verify the effectiveness of the proposed method, the improved network is trained and tested, and the test results show that the improved network training model can accurately identify and mark the carbon marks on the pressboard surface.

Deep learning techniques‐based perfection of multi‐sensor fusion oriented human‐robot interaction system for identification of dense organisms

For detection of dense small-target organisms with indistinct features in complex background, the efficiency and accuracy of traditional target detection methods are low. Multi-sensor fusion oriented human-robot interaction (HRI) system has facilitated biologists to process and analyse data. For this, several deep learning models based on convolutional neural network (CNN) are improved and compared to study the species and density of dense organisms in deep-sea hydrothermal vent, which are fused it with related environmental information given by position sensors and conductivity-temperature-depth (CTD) sensors, so as to perfect multi-sensor fusion oriented HRI system. Firstly, the authors combined different meta-architectures and different feature extractors, and obtained five object identification algorithms based on CNN. Then, they compared computational cost of feature extractors and weighed the pros and cons of each algorithm from mean detection speed, correlation coefficient and mean class-specific confidence score to confirm that Faster Region-based CNN (R-CNN)_InceptionNet is the best algorithm applicable to hydrothermal vent biological dataset. Finally, they calculated the cognitive accuracy of rimicaris exoculata in dense and sparse areas, which were 88.3% and 95.9% respectively, to analyse the performance of the Faster R-CNN_InceptionNet. Results show that the proposed method can be used in the multi-sensor fusion oriented HRI system for the statistics of dense organisms in complex environments.

Improved YOLOv4 Marine Target Detection Combined with CBAM

Marine target detection technology plays an important role in sea surface monitoring, sea area management, ship collision avoidance, and other fields. Traditional marine target detection algorithms cannot meet the requirements of accuracy and speed. This article uses the advantages of deep learning in big data feature learning to propose the YOLOv4 marine target detection method fused with a convolutional attention module. Marine target detection datasets were collected and produced and marine targets were divided into ten categories, including speedboat, warship, passenger ship, cargo ship, sailboat, tugboat, and kayak. Aiming at the problem of insufficient detection accuracy of YOLOv4’s self-built marine target dataset, a convolutional attention module is added to the YOLOv4 network to increase the weight of useful features while suppressing the weight of invalid features to improve detection accuracy. The experimental results show that the improved YOLOv4 has higher detection accuracy than the original YOLOv4, and has better detection results for small targets, multiple targets, and overlapping targets. The detection speed meets the real-time requirements, verifying the effectiveness of the improved algorithm.

Object Detection Algorithm Based on YOLOv3 Model to Detect Occluded Targets

Target detection has experienced a rapid development stage from the concept to the practical application, the core technology has been broken through, many difficult problems have been solved, and it has been widely used in various fields. However, the existing target detection models still have some technical problems such as incomplete detection and low accuracy. Therefore, this paper proposes a target detection algorithm based on V3 model to detect occluded targets. This paper makes an in-depth investigation and Research on the existing target detection model, analyzes the shortcomings of the current traditional target detection model through the test results, and the advantages of the network detection model based on YOLOv3 in target detection. Aiming at the main problems existing in the current detection technology, this paper puts forward the optimization and improvement scheme. By reconstructing the whole framework of multi-target tracking algorithm, optimizing the feature fusion algorithm, and improving the NMS algorithm, the scheme greatly improves the accuracy of detecting the occluded objects. The simulation results show that the improved scheme can improve the accuracy by 12.3% compared with the traditional scheme. The analysis shows that the target detection model based on YOLOv3 network can effectively improve the detection accuracy of the model.

Robot grasping method optimization using improved deep deterministic policy gradient algorithm of deep reinforcement learning

Robot grasping has become a very hot research field so that the requirements for robot operation are getting higher and higher. In previous research studies, the use of traditional target detection algorithms for grasping is often very inefficient, and this article is dedicated to improving the deep reinforcement learning algorithm to improve the grasping efficiency and solve the problem of robots dealing with the impact of unknown disturbances on grasping. Using the characteristic that deep reinforcement learning actively explores the unknown environment, a Gaussian parameter Deep Deterministic Policy Gradient (Gaussian-DDPG) algorithm based on the Importance-Weighted Autoencoder (IWAE) is proposed to realize the robot's autonomous learning of the grasping task. Traditional coordinate positioning methods and deep learning methods have poor grasping effects for disturbed situations (such as the movement of the target object). The IWAE algorithm is used to compress the high-dimensional information of the original visual input to the hidden space and pass it to the deep reinforcement learning network as part of the state value. Based on the classic DDPG algorithm, it smoothly adds Gaussian parameters to improve the exploratory nature of the algorithm, dynamically sets the robot grasping space parameters to adapt to the workspace of multiple scales, and finally, realizes the accurate grasping of the robot. Relying on the possible position information deviation of the visual information, the control of the grasping position by the manipulator torque information is further optimized to improve the grasping efficiency of disturbed objects.

RISTDnet: Robust Infrared Small Target Detection Network

The infrared (IR) small target detection algorithm with a high detection rate, low false alarm rate, and high real-time performance has significant application value in the field of IR remote sensing. IR small targets in complex backgrounds have low contrast and low signal-to-noise ratio (SNR). Therefore, small target detection is more difficult. Traditional IR small target detection is generally implemented by local contrast methods (LCM), nonlocal autocorrelation methods (NAM), and adaptive segmentation. In this letter, a robust infrared small target detection network (RISTDnet) is proposed based on deep learning. In RISTDnet, a feature extraction framework combining handcrafted feature methods and convolutional neural networks is constructed, a mapping network between feature maps and the likelihood of small targets in the image is established, and a threshold is applied on the likelihood map to segment real targets. Experimental results show that the RISTDnet can detect small targets with different sizes and low SNRs in complex backgrounds and have better effectiveness and robustness against existing algorithms.