Towards Pedestrian Target Detection with Optimized Mask R-CNN

Aiming at the problem of low pedestrian target detection accuracy, we propose a detection algorithm based on optimized Mask R-CNN which uses the latest research results of deep learning to improve the accuracy and speed of detection results. Due to the influence of illumination, posture, background, and other factors on the human target in the natural scene image, the complexity of target information is high. SKNet is used to replace the part of the convolution module in the depth residual network model in order to extract features better so that the model can adaptively select the best convolution kernel during training. In addition, according to the statistical law, the length-width ratio of the anchor box is modified to make it more accord with the natural characteristics of the pedestrian target. Finally, a pedestrian target dataset is established by selecting suitable pedestrian images in the COCO dataset and expanded by adding noise and median filtering. The optimized algorithm is compared with the original algorithm and several other mainstream target detection algorithms on the dataset; the experimental results show that the detection accuracy and detection speed of the optimized algorithm are improved, and its detection accuracy is better than other mainstream target detection algorithms.

Download Full-text

A New Steel Defect Detection Algorithm Based on Deep Learning

Computational Intelligence and Neuroscience ◽

10.1155/2021/5592878 ◽

2021 ◽

Vol 2021 ◽

pp. 1-13

Author(s):

Weidong Zhao ◽

Feng Chen ◽

Hancheng Huang ◽

Dan Li ◽

Wei Cheng

Keyword(s):

Deep Learning ◽

Target Detection ◽

Defect Detection ◽

Steel Surface ◽

Surface Defect ◽

Detection Algorithm ◽

Detection Accuracy ◽

Average Precision ◽

Original Algorithm ◽

Surface Defect Detection

In recent years, more and more scholars devoted themselves to the research of the target detection algorithm due to the continuous development of deep learning. Among them, the detection and recognition of small and complex targets are still a problem to be solved. The authors of this article have understood the shortcomings of the deep learning detection algorithm in detecting small and complex defect targets and would like to share a new improved target detection algorithm in steel surface defect detection. The steel surface defects will affect the quality of steel seriously. We find that most of the current detection algorithms for NEU-DET dataset detection accuracy are low, so we choose to verify a steel surface defect detection algorithm based on machine vision on this dataset for the problem of defect detection in steel production. A series of improvement measures are carried out in the traditional Faster R-CNN algorithm, such as reconstructing the network structure of Faster R-CNN. Based on the small features of the target, we train the network with multiscale fusion. For the complex features of the target, we replace part of the conventional convolution network with a deformable convolution network. The experimental results show that the deep learning network model trained by the proposed method has good detection performance, and the mean average precision is 0.752, which is 0.128 higher than the original algorithm. Among them, the average precision of crazing, inclusion, patches, pitted surface, rolled in scale and scratches is 0.501, 0.791, 0.792, 0.874, 0.649, and 0.905, respectively. The detection method is able to identify small target defects on the steel surface effectively, which can provide a reference for the automatic detection of steel defects.

Download Full-text

Multi-Scale Safety Helmet Detection Based on SAS-YOLOv3-Tiny

Applied Sciences ◽

10.3390/app11083652 ◽

2021 ◽

Vol 11 (8) ◽

pp. 3652

Author(s):

Rao Cheng ◽

Xiaowei He ◽

Zhonglong Zheng ◽

Zhentao Wang

Keyword(s):

Detection Algorithm ◽

Model Complexity ◽

Detection Accuracy ◽

Global Features ◽

Original Algorithm ◽

Scale Feature ◽

Detection Algorithms ◽

Multi Scale ◽

Local And Global Features ◽

A Minor

In the practical application scenarios of safety helmet detection, the lightweight algorithm You Only Look Once (YOLO) v3-tiny is easy to be deployed in embedded devices because its number of parameters is small. However, its detection accuracy is relatively low, which is why it is not suitable for detecting multi-scale safety helmets. The safety helmet detection algorithm (named SAS-YOLOv3-tiny) is proposed in this paper to balance detection accuracy and model complexity. A light Sandglass-Residual (SR) module based on depthwise separable convolution and channel attention mechanism is constructed to replace the original convolution layer, and the convolution layer of stride two is used to replace the max-pooling layer for obtaining more informative features and promoting detection performance while reducing the number of parameters and computation. Instead of two-scale feature prediction, three-scale feature prediction is used here to improve the detection effect about small objects further. In addition, an improved spatial pyramid pooling (SPP) module is added to the feature extraction network to extract local and global features with rich semantic information. Complete-Intersection over Union (CIoU) loss is also introduced in this paper to improve the loss function for promoting positioning accuracy. The results on the self-built helmet dataset show that the improved algorithm is superior to the original algorithm. Compared with the original YOLOv3-tiny, the SAS-YOLOv3-tiny has significantly improved all metrics (including Precision (P), Recall (R), Mean Average Precision (mAP), F1) at the expense of only a minor speed while keeping fewer parameters and amounts of calculation. Meanwhile, the SAS-YOLOv3-tiny algorithm shows advantages in accuracy compared with lightweight object detection algorithms, and its speed is faster than the heavyweight model.

Download Full-text

Object Detection from the Video Taken by Drone via Convolutional Neural Networks

Mathematical Problems in Engineering ◽

10.1155/2020/4013647 ◽

2020 ◽

Vol 2020 ◽

pp. 1-10 ◽

Cited By ~ 1

Author(s):

Chenfan Sun ◽

Wei Zhan ◽

Jinhiu She ◽

Yangyang Zhang

Keyword(s):

Object Detection ◽

Target Detection ◽

Image Data ◽

Detection Algorithm ◽

Detection Accuracy ◽

Detection Algorithms ◽

Detection Systems ◽

Video Frames ◽

Performance Effects ◽

And Performance

The aim of this research is to show the implementation of object detection on drone videos using TensorFlow object detection API. The function of the research is the recognition effect and performance of the popular target detection algorithm and feature extractor for recognizing people, trees, cars, and buildings from real-world video frames taken by drones. The study found that using different target detection algorithms on the “normal” image (an ordinary camera) has different performance effects on the number of instances, detection accuracy, and performance consumption of the target and the application of the algorithm to the image data acquired by the drone is different. Object detection is a key part of the realization of any robot’s complete autonomy, while unmanned aerial vehicles (UAVs) are a very active area of this field. In order to explore the performance of the most advanced target detection algorithm in the image data captured by UAV, we have done a lot of experiments to solve our functional problems and compared two different types of representative of the most advanced convolution target detection systems, such as SSD and Faster R-CNN, with MobileNet, GoogleNet/Inception, and ResNet50 base feature extractors.

Download Full-text

Research on Lightweight Infrared Pedestrian Detection Model Algorithm for Embedded Platform

Security and Communication Networks ◽

10.1155/2021/1549772 ◽

2021 ◽

Vol 2021 ◽

pp. 1-7

Author(s):

Zhaoli Wu ◽

Xin Wang ◽

Chao Chen

Keyword(s):

Real Time ◽

Target Detection ◽

Pedestrian Detection ◽

Infrared Image ◽

Far Infrared ◽

Detection Algorithm ◽

Model Parameters ◽

Detection Accuracy ◽

Detection Model ◽

Embedded Platform

Due to the limitation of energy consumption and power consumption, the embedded platform cannot meet the real-time requirements of the far-infrared image pedestrian detection algorithm. To solve this problem, this paper proposes a new real-time infrared pedestrian detection algorithm (RepVGG-YOLOv4, Rep-YOLO), which uses RepVGG to reconstruct the YOLOv4 backbone network, reduces the amount of model parameters and calculations, and improves the speed of target detection; using space spatial pyramid pooling (SPP) obtains different receptive field information to improve the accuracy of model detection; using the channel pruning compression method reduces redundant parameters, model size, and computational complexity. The experimental results show that compared with the YOLOv4 target detection algorithm, the Rep-YOLO algorithm reduces the model volume by 90%, the floating-point calculation is reduced by 93.4%, the reasoning speed is increased by 4 times, and the model detection accuracy after compression reaches 93.25%.

Download Full-text

Study of BLAST Signal Detection Algorithm in LTE System

Advanced Materials Research ◽

10.4028/www.scientific.net/amr.756-759.3183 ◽

2013 ◽

Vol 756-759 ◽

pp. 3183-3188

Author(s):

Tao Lei ◽

Deng Ping He ◽

Fang Tang Chen

Keyword(s):

Signal Detection ◽

High Speed ◽

Data Communication ◽

Detection Algorithm ◽

Detection Performance ◽

Detection Algorithms ◽

High Speed Data ◽

Simulation Results ◽

Better Than

BLAST can achieve high speed data communication. Its signal detection directly affects performance of BLAST receiver. This paper introduced several signal detection algorithmsZF algorithm, MMSE algorithm, ZF-SIC algorithm and MMSE-SIC algorithm. The simulation results show that the traditional ZF algorithm has the worst performance, the traditional MMSE algorithm and the ZF-SIC algorithm is similar, but with the increase of the SNR, the performance of ZF-SIC algorithm is better than MMSE algorithm. MMSE-SIC algorithm has the best detection performance in these detection algorithms.

Download Full-text

Angle Distance-Based Hierarchical Background Separation Method for Hyperspectral Imagery Target Detection

Remote Sensing ◽

10.3390/rs12040697 ◽

2020 ◽

Vol 12 (4) ◽

pp. 697 ◽

Cited By ~ 1

Author(s):

Xiaohui Hao ◽

Yiquan Wu ◽

Peng Wang

Keyword(s):

Target Detection ◽

Hyperspectral Imagery ◽

Original Data ◽

Separation Method ◽

Detection Accuracy ◽

Target Information ◽

Detection Algorithms ◽

Target Spectrum ◽

Candidate Target ◽

The Given

Traditional detectors for hyperspectral imagery (HSI) target detection (TD) output the result after processing the HSI only once. However, using the prior target information only once is not sufficient, as it causes the inaccuracy of target extraction or the unclean separation of the background. In this paper, the target pixels are located by a hierarchical background separation method, which explores the relationship between the target and the background for making better use of the prior target information more than one time. In each layer, there is an angle distance (AD) between each pixel spectrum in HSI and the given prior target spectrum. The AD between the prior target spectrum and candidate target ones is smaller than that of the background pixels. The AD metric is utilized to adjust the values of pixels in each layer to gradually increase the separability of the background and the target. For making better discrimination, the AD is calculated through the whitened data rather than the original data. Besides, an elegant and ingenious smoothing processing operation is employed to mitigate the influence of spectral variability, which is beneficial for the detection accuracy. The experimental results of three real hyperspectral images show that the proposed method outperforms other classical and recently proposed HSI target detection algorithms.

Download Full-text

Multirapid Serial Visual Presentation Framework for EEG-Based Target Detection

BioMed Research International ◽

10.1155/2017/2049094 ◽

2017 ◽

Vol 2017 ◽

pp. 1-12 ◽

Cited By ~ 7

Author(s):

Zhimin Lin ◽

Ying Zeng ◽

Hui Gao ◽

Li Tong ◽

Chi Zhang ◽

...

Keyword(s):

Image Retrieval ◽

Target Detection ◽

Visual Presentation ◽

Research Direction ◽

Detection Methods ◽

Detection Accuracy ◽

Single Trial ◽

Target Image ◽

Detection Algorithms ◽

New Research

Target image detection based on a rapid serial visual presentation (RSVP) paradigm is a typical brain-computer interface system with various applications, such as image retrieval. In an RSVP paradigm, a P300 component is detected to determine target images. This strategy requires high-precision single-trial P300 detection methods. However, the performance of single-trial detection methods is relatively lower than that of multitrial P300 detection methods. Image retrieval based on multitrial P300 is a new research direction. In this paper, we propose a triple-RSVP paradigm with three images being presented simultaneously and a target image appearing three times. Thus, multitrial P300 classification methods can be used to improve detection accuracy. In this study, these mechanisms were extended and validated, and the characteristics of the multi-RSVP framework were further explored. Two different P300 detection algorithms were also utilized in multi-RSVP to demonstrate that the scheme is universally applicable. Results revealed that the detection accuracy of the multi-RSVP paradigm was higher than that of the standard RSVP paradigm. The results validate the effectiveness of the proposed method, and this method can provide a whole new idea in the field of EEG-based target detection.

Download Full-text

Image target detection algorithm compression and pruning based on neural network

Computer Science and Information Systems ◽

10.2298/csis200316007s ◽

2021 ◽

Vol 18 (2) ◽

pp. 499-516

Author(s):

Yan Sun ◽

Zheping Yan

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Target Detection ◽

Network Model ◽

Detection Algorithm ◽

Image Feature ◽

Propagation Mode ◽

Learning Technology ◽

Image Detection ◽

Detection Algorithms

The main purpose of target detection is to identify and locate targets from still images or video sequences. It is one of the key tasks in the field of computer vision. With the continuous breakthrough of deep machine learning technology, especially the convolutional neural network model shows strong Ability to extract image feature in the field of digital image processing. Although the model research of target detection based on convolutional neural network is developing rapidly, but there are still some problems in practical applications. For example, a large number of parameters requires high storage and computational costs in detected model. Therefore, this paper optimizes and compresses some algorithms by using early image detection algorithms and image detection algorithms based on convolutional neural networks. After training and learning, there will appear forward propagation mode in the application of CNN network model, providing the model for image feature extraction, integration processing and feature mapping. The use of back propagation makes the CNN network model have the ability to optimize learning and compressed algorithm. Then research discuss the Faster-RCNN algorithm and the YOLO algorithm. Aiming at the problem of the candidate frame is not significant which extracted in the Faster- RCNN algorithm, a target detection model based on the Significant area recommendation network is proposed. The weight of the feature map is calculated by the model, which enhances the saliency of the feature and reduces the background interference. Experiments show that the image detection algorithm based on compressed neural network image has certain feasibility.

Download Full-text

Small Target Detection Algorithm Based on Transfer Learning and Deep Separable Network

Journal of Sensors ◽

10.1155/2021/9006288 ◽

2021 ◽

Vol 2021 ◽

pp. 1-10

Author(s):

Peng Wang ◽

Haiyan Wang ◽

Xiaoyan Li ◽

Lingling Zhang ◽

Ruohai Di ◽

...

Keyword(s):

Transfer Learning ◽

Target Detection ◽

Activation Function ◽

Detection Algorithm ◽

Detection Accuracy ◽

Small Target ◽

Data Set ◽

Small Target Detection ◽

Low Level ◽

Half Axis

With the development of deep learning, target detection from vision sensor has achieved high accuracy and efficiency. However, small target detection remains a challenge due to inadequate use of semantic information and detailed texture information of underlying features. To solve the above problems, this paper proposes a small target detection algorithm based on Mask R-CNN model which integrates transfer learning and deep separable network. Firstly, the feature pyramid fusion structure is introduced to enhance the learning effect of low-level and high-level features, especially to strengthen the information channel of low-level feature and meanwhile optimize the feature information of small target. Secondly, the ELU function is used as the activation function to solve the problem that the original activation function disappears in the negative half axis gradient. Finally, a new loss function F-Softmax combined with Focal Loss was adopted to solve the imbalance of positive and negative sample proportions. In this paper, self-made data set is used to carry out experiments, and the experimental results show that the proposed algorithm makes the detection accuracy of small targets reach 66.5%.

Download Full-text

TWO-LEAD JOINT DETECTION ALGORITHM OF QRS COMPLEXES IN ECG SIGNAL

Journal of Mechanics in Medicine and Biology ◽

10.1142/s0219519421400728 ◽

2021 ◽

Author(s):

WANSONG XU ◽

TIANWU CHEN ◽

FANYU DU

Keyword(s):

Detection Algorithm ◽

Detection Accuracy ◽

Ecg Signal ◽

Joint Detection ◽

Ecg Signals ◽

Detection Algorithms ◽

Threshold Adjustment ◽

Current Algorithm ◽

Electrocardiogram Ecg

Objective: The detection of QRS complexes is an important part of computer-aided analysis of electrocardiogram (ECG). However, most of the existing detection algorithms are mainly for single-lead ECG signals, which requires high quality of signal. If the signal quality decreases suddenly due to some interference, then the current algorithm is easy to cause misjudgment or missed detection. To improve the detection ability of QRS complexes under sudden interference, we study the QRS complexes information on multiple leads in-depth, and propose a two-lead joint detection algorithm of QRS complexes. Methods: Firstly, the suspected QRS complexes are screened on the main lead. For the suspected QRS complexes with low confidence and the complexes that may be missed, further accurate detection and joint judgment shall be carried out at the corresponding position of the auxiliary lead. At the same time, the adaptive threshold adjustment algorithm and backtracking mechanism are used to modify the detection results. Results: The proposed detection algorithm is validated using 48 ECG records of the MIT-BIH arrhythmia database, and achieves average detection accuracy of 99.71%, sensitivity of 99.88% and positive predictivity of 99.81%. Conclusion: The proposed algorithm has high accuracy, which can effectively deal with the sudden interference of ECG signal. Meanwhile, the algorithm requires small amount of computation, and can be embedded into hardware for real-time detection.

Download Full-text