Faculty Profiles - ZHANG Lifeng

Papers - ZHANG Lifeng

Division display41 - 60 of about 223／ All the affair displays >>

Research on Marine Life Detection Algorithm Based on Deep Learning Reviewed International journal

Yuting Wang, Seiichi Serikawa, Wenwen Shen, Lifeng Zhang, Zijun Yang, Yuliang Gao, Aoran Xi

Proceedings of the 9th IIAE International Conference on Intelligent Systems and Image Processings 2022 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 133 - 138 2022.09

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kobe 2022.09.06 - 2022.09.10

Marine life detection based on deep learning has always been the focus and difficulty in the field of deep learning target detection. There are two main problems. First, underwater imaging often has problems such as low image contrast, blurred image, color shift, and blue-green tint of the background. How to solve these problems in imaging. Second, there are many algorithms in the field of deep learning target detection, how to choose a suitable algorithm for marine life detection. Aiming at the above problems, this paper firstly introduces three algorithms suitable for underwater image enhancement: the method based on dark channel prior, the method based on histogram equalization and the method based on Retinex. Then, six commonly used deep learning target detection algorithms are introduced from two aspects: target detection algorithm based on candidate region proposal and target detection algorithm based on regression: R-CNN, Fast R-CNN, Faster R-CNN, Mask R-CNN, YOLO, SSD, and each analyzes their advantages and disadvantages. It provides a theoretical and method reference for deep learning-based marine biological detection.

DOI： 10.12792/icisip2022.023

Other Link： https://www2.ia-engineers.org/conference/index.php/icisip/icisip2022/paper/view/2633

Design of Control System for Vegetable Soilless Culture Reviewed International journal

Yaoyao Chen, Zijun Yang, Shi Zhou, Aoran Xi, Yuting Wang, Zhang Lifeng

Proceedings of the 9th IIAE International Conference on Intelligent Systems and Image Processings 2022 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 146 - 154 2022.09

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

Japan Kobe 2022.09.06 - 2022.09.10

This design combines the automation control technology with the latest soilless cultivation technology to make the vegetable transparent and controllable in all aspects of the whole growth cycle. It is different from the current technology. It is more innovative to modularize the cultivation unit and monitoring unit , so that its application scope is greatly expanded. Under certain conditions, it can also be added as a subsystem of other systems. With application, based on array distributed sensors and algorithm, the atomization of nutrient solution is controlled to optimize the nutrient conditions and make the growth state of vegetables best. Considering the economy and practicability of the actual construction of the system, Proteus and the C language is used to simulate the circuits in the design process, construction of the project fully ensure the reliability of the system and the transplant of the multi-platform environment.

DOI： 10.12792/icisip2022.025

Other Link： https://www2.ia-engineers.org/conference/index.php/icisip/icisip2022/paper/view/2638

Identifying User's Smartphone Possession Location and Behavior Including Untrained Classes by Using Deep Metric Learning Reviewed

Rui Kitahara, Lifeng Zhang

Proceedings of the 10th IIAE International Conference on Industrial Application Engineering 2022 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 213 - 219 2022.03

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

Japan Matsue 2022.03.26 - 2022.03.31

In our previous study,(1) we confirmed that it was possible to classify the user’s smartphone possession location and behavior simultaneously using Mdk-ResNet.(2) However, there was a problem that when untrained classes were given as input, it was impossible to output them as unknown classes using normal classifications as in our and other’s previous studies. Therefore, this time, we aimed to output untrained data as an unknown class by learning the distance between images using deep metric learning, and not only classifying the learned classes. In this study, we used a smartphone’s accelerometer to acquire the data of experimental participants and created a trained model by learning through deep metric learning. We classified the user’s smartphone possession location and behavior, including the unknown class of unlearned data, by calculating the cosine similarity between the image vector prepared in advance and the image vector output from the trained model during inference using the trained model. As a result, we confirm not only that the untrained data can be output as the unknown class, but also that the accuracy is comparable to the previous study.

DOI： 10.12792/iciae2022.039

Other Link： https://www2.ia-engineers.org/conference/index.php/iciae/iciae2022/paper/view/2604

Research Status of Fault Diagnosis of Large and Medium-sized Vertical Water Pump Units Reviewed International journal

Aoran Xi, Lifeng Zhang, Yueping Mo

Proceedings of the 10th IIAE International Conference on Industrial Application Engineering 2022 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 156 - 161 2022.03

Language：English Publishing type：Research paper (international conference proceedings)

Japan Matsue 2022.03.26 - 2022.03.31

This paper analyzes the development process and status quo of the fault diagnosis technology of vertical water pump units. Several common fault detection methods of vertical water pump units are introduced in detail from the analytical model-based processing method, signal-based processing method, and knowledge-based processing method. The advantages and disadvantages are analyzed and summarized, providing theoretical methods for fault diagnosis of large and medium-sized vertical water pump units.

DOI： 10.12792/iciae2022.028

Other Link： https://www2.ia-engineers.org/conference/index.php/iciae/iciae2022/paper/view/2598

An Adaptive Threshold Canny Edge Detection Method for Applications with Simple Background Reviewed

Shi Zhou, Zijun Yang, Zhen Li, Yuliang Gao, Seiichi Serikawa, Mitsunori Mizumachi, Lifeng Zhang

Proceedings of the 10th IIAE International Conference on Industrial Application Engineering 2022 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 123 - 128 2022.03

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

Japan Matsue 2022.03.26 - 2022.03.31

Edge detection is an important image processing problem with many practical applications in different fields. While many methods have been proposed, further improvements still need to be made for specific applications. This paper focuses on the application of Canny edge in the image with a simple background, proposed an adaptive threshold Canny edge detection method. Considering the distribution of the object and the solid-color background, a pixels intensity statistic is operated for the corresponding gray image. In this way, the double thresholds are determined by analyzing the statistical result. Furthermore, to verify the effectiveness of the proposed method, experiments are done on images that are taken with different conditions. The result shows that Canny edge can be well detected by utilizing the two adaptive threshold values and have strong robustness to different conditions. Particularly in the applications, such as object detection and image segmentation, the adaptive threshold Canny edge detection method plays a meaningful role.

DOI： 10.12792/iciae2022.023

Other Link： https://www2.ia-engineers.org/conference/index.php/iciae/iciae2022/paper/view/2597

Person Re-identification for Real-time Video Surveillance Based on Deep Learning Reviewed International journal

Miaomiao Zhu, Shengrong Gong, Zhenjiang Qian, Seiichi Serikawa, Lifeng Zhang

Proceedings of the Joint Symposium of The Twenty-Seventh International Symposium on Artificial Life and Robotics (AROB 27th 2022) ( International Society of Artificial Life and Robotics ) 98 - 103 2022.01

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

Person re-identification (ReID), as an instance-level recognition problem, aims at automatically retrieving a person- of-interest across multiple non-overlapping camera views, which is considered as a sub problem of image retrieval. Unlike the traditional ReID research, the proposed system aims to combine pedestrian detection and ReID to perform one-step pedestrian detection and search. For real-time video surveillance obtained by cameras distributed in different locations, given a person-of- interest to be queried, the goal of our method is to search a person from the whole scene images or videos directly instead of matching them with manually cropped pedestrians in the existing datasets. To further verify the effectiveness of the proposed method in the real scene，many experimental datasets of complex scenes are collected respectively. Finally, the experimental results show that our method is effective in commonly used datasets and has achieved good results for real-time video surveillance in the real scene.

Identification of Rice Disease with Shuffle Attention & Inception Reviewed International journal

Yuliang Gao1, Zhen Li, Lifeng Zhang, Seiichi Serikawa, Xiangying Xu, Bin Li

Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

This paper presents an identification algorithm of rice diseases and insect pests based on a convolutional neural net- work. It aims to solve the problems of a large number of parameters and the low accuracy of the traditional algorithm. The Inception-ResNetV1 module and the shuffle attention module are combined. It achieves the effects of significantly reducing parameters and improving accuracy. And this model is trained by SAM optimizer to get a better result. Compared with tradi- tional models ResNet50 & Inception-ResNetV1, the highest accuracy reaches 95.66%, the parameters in the self-made dataset decreased by 47.6% and 34.02%, and the accuracy increased by 1.91% and 1.32%, respectively.

Proposal for a system for visually impaired people to identify products in stores Reviewed International journal

Tomoyuki Zaizen, Rui Kitahara, Lifeng Zhang

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

In recent years, Japan has been facing the problem of an aging population due to the aging of society. In addition, many elderly people are visually impaired, and the number of visually impaired people is expected to increase in the future. However, the aging society is also facing a shortage of caregivers, as the elderly must care for the elderly. Therefore, the problem of not having enough people to care for the increasing number of visually impaired people will arise in Japan in the future. The shortage of caregivers may lead to the tendency of the elderly visually impaired to refrain from going out. Therefore, a system that allows the visually impaired to live without care is needed to prevent this.
Prior work includes a system that identifies products using image processing techniques, namely motion detection and image classification using machine learning techniques. However, in the system of this previous research, the number of products that can be identified is as few as six, and the product identification rate is not sufficiently high.
Therefore, in this research, we attempted to increase the number of products that can be identified and also to improve the identification rate of the products.

The Hilbert Curve-based Speech Signal Visualization for Speech Emotion Recognition Reviewed International journal

Zijun Yang, Lifeng Zhang, Shi Zhou, Seiichi Serikawa

Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

This paper proposed a method to convert the one-dimensional speech signal into two-dimensional image signal, so that the mature image processing skills can be used to analysis speech signal and implement speech emotion recognition. The article uses the Hilbert Curve which can fill the entire space and give a mapping between one-dimensional and two-dimensional space. Moreover, to better illuminate the effectiveness of the proposed method, we use the proposed method convert the six emotion language signals dataset into the images, and a neural network is trained on the created dataset. The result shows that the trained model can achieve an accuracy rate of 51.7%. It means use the method that convert the speech signal into image signal implement speech emotion recognition is well. In the follow-up, we will further optimize the algorithm to improve the accuracy.

Behavioral Recognition Considering Multiple Possession Positions Using Smartphone Accelerometers by Transfer Learning Reviewed International journal

Rui Kitahara, Lifeng Zhang

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

With the widespread use of smartphones and wearable devices in recent years, various applications have been devel- oped to use their built-in sensors.Among them, HAR (Human Activity Recognition) using accelerometers has the advantage of privacy protection compared to camera-based data acquisition.Therefore, it is thought a good approach to watch over the elderly who live alone in the coming elder society.However, most of the behavior recognition has been done by fixing the possession position of the smartphone in one place, measuring the data, and recognizing the behavior.However, this is not according to the actual situation because people carry smartphones around in different places.Therefore, the purpose of this study is to recognize behavior by considering various possession positions. In this study, we set up four different positions for smartphone posses- sion. And we set up three different behavior. By combining all of the possession positions (4 types) and behaviors (3 types), we created a data set with 12 classes for classification. After data measurement, the measured time-series data were converted into 2D images using the GAF[1] algorithm as a preprocessing step.Then, the time-series data, which became 2D images, were trained with Mdk-ResNet[2] for classification. The dataset created in this study has a smaller number of datasets compared to previous studies.Therefore, in this study, we further aimed to improve the accuracy by also performing transfer learning.In this study, we created the Mdk-ResNet trained model using the WISDM dataset, a large-scale behavioral dataset frequently used in behavioral recognition research, and performed transfer learning.As a result, the results were Accuracy:0.9899, F1:0.9890, Recall:0.9895, and Precision:0.9887, when not using the transfer learning. This result indicates that recognition was possible with relatively high accuracy.In addition, by using transfer learning, we could obtain Accuracy:0.9921,F1:0.9913,Recall:0.9913 and Precision:0.9905, which was even higher recognition accuracy.

A Keypoint-Based and Anchor-Free 3D Detection for Autonomous Driving Using Monocular Camera Reviewed International journal

Zhen Li, Yuliang Gao, Qingqing Hong, Yuren Du, Seiichi Serikawa and Lifeng Zhang

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

Autonomous driving technology has received great attention from scientific research and social marketing. 3D object detection technology is the core research topic of autonomous driving. Nowadays, most 3D object detection researches heavily depend on lidar and radar sensors, which could directly offer depth information. However, inspired by human driving experiences, 3D world perception can be obtained by pure visual scheme. This paper proposed an anchor-free and keypoint-based 3D object detection framework using a monocular camera. Our work took CenterNet as baseline and made several improvements to achieve better 3D detection performance. First, we utilized the rotation angle regression method in polar coordinate system, which has simplified the original framework greatly. Then, we proposed the ellipse Gaussian Filter method to fit the ground truth rectangular boxes much more better. Finally, this paper also improved the backbone network of DLA-34, the more complex neural network performances better in complex 3D detection task. Our proposed framework can improve 3D detection by 6.9% in pedestrian classification. And the whole detection performance, including 2D detection, also improved to 97.8% for vehicles.

Self-Supervised Visual Odometry Based on the Projection of Reliable Pixels Reviewed International journal

Shi Zhou, Zijun Yang, He Li, Seiichi Serikawa, Mitsunori Mizumachi and Lifeng Zhang

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings)

日本別府 2022.01.25 - 2022.01.27

Complex outdoor environment includes variable lighting, moving objects and direct sunlight, visual odometry in outdoor environments remains a challenging problem. In this work, a self-supervised pose estimation network is trained based on the projection between adjacent frames. The pose between two adjacent frames is a 6D vector. It can be solved with a part of effective pixels instead of all the pixels. Thus, only reliable pixels are under consideration when training. To eliminating the influence of unreliable pixels, outside mask and poor projection mask are proposed to check them out. Furthermore, we evaluated the proposed method on the KITTI datasets and compared it with other deep learning-based visual odometry methods. The result shows that the proposed method can detect poor projection and outside pixels well. Moreover, it achieves higher accuracy pose estimation.

Higher accuracy self-supervised visual odometry with reliable projection Reviewed International journal

Zhou S., Yang Z., Zhu M., Li H., Serikawa S., Mizumachi M., Zhang L.

Artificial Life and Robotics 2022.01

A survey of 3D object detection algorithms for intelligent vehicles development Reviewed International journal

Zhen Li, Yuren Du, Miaomiao Zhu, Shi Zhou & Lifeng Zhang

Artificial Life and Robotics ( Springer ) 27 ( 1 ) 115 - 122 2021.11

Authorship：Last author Language：English Publishing type：Research paper (scientific journal)

With the rapid development of Artificial Intelligent algorithms on Computer Vision, 2D object detection has greatly succeeded and been applied in various industrial products. In the past several years, the accuracy of 2D object detection has been dramatically improved, even beyond the human eyes detection ability. However, there is still a limitation of 2D object detection for the applications of Intelligent Driving. A safe and reliable self-driving car needs to detect a 3D model of the around objects so that an intelligent driving car has a perception ability to real driving situations. This paper systematically surveys the development of 3D object detection methods applied to intelligent driving technology. This paper also analyzes the shortcomings of the existing 3D detection algorithms and the future development directions of 3D detection algorithms on intelligent driving.

DOI： 10.1007/s10015-021-00711-0

Scopus

Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85118385846&origin=inward

A 3D Object Detection Framework for Intelligent Driving using YOLOv4 Reviewed International journal

Zhen Li, Yuren Du, Qingqing Hong, Seiichi Serikawa, Lifeng Zhang

Proceedings of the 8th IIAE International Conference on Intelligent Systems and Image Processings 2021 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 9 - 15 2021.09

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kobe 2021.09.06 - 2021.09.10

As the continuous evolution of the vehicles' industry, the self-driving technology has become an indispensable part in today's intelligent vehicles. While 3D object detection is the most core technique in the self-driving technology. A little improvement in the accuracy and efficiency of a 3D object detector will be of the utmost importance to intelligent driving and human life. So this paper focused on the improvement of detecting accuracy based on the yolov4, which is also an effective object detection network. This work transformed yolov4 from a 2D detector into a 3D detector, and also improved the original yolov4 neural network with the CSPDarknet53 backbone network. The proposed framework has a more effective and accurate performance in the 3D object detecting for intelligent driving.

DOI： 10.12792/icisip2021.004

Other Link： https://www2.ia-engineers.org/conference/index.php/icisip/icisip2021/paper/view/2495

Behavioral Estimation for Multiple Possession Positions Using Smartphone Accelerometers Reviewed International journal

Rui Kitahara, Lifeng Zhang

Proceedings of the 8th IIAE International Conference on Intelligent Systems and Image Processings 2021 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 171 - 174 2021.09

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kobe 2021.09.06 - 2021.09.10

In recent years, with the widespread use of smartphones and wearable devices, a variety of research has been conducted using built-in sensors. For example, height estimation and road condition estimation have been conducted. In addition, behavioral estimation of the smartphone holder, possession position estimation, and person estimation have also been conducted. However, most of them measure data by fixing the possession position at a single location, and do not take the possession position into account when estimating behavior. In this research, we aim to estimate the behavior of a person by considering multiple possession positions, because it is necessary to estimate the behavior of a person by considering various possession positions when actually using behavior estimation as a system. In addition, by treating the time series data acquired by the 3-axis acceleration sensor of a smartphone as a 2-dimensional image using the GAF algorithm,(1) class classification by machine learning is performed.

DOI： 10.12792/icisip2021.030

Other Link： https://www2.ia-engineers.org/conference/index.php/icisip/icisip2021/paper/view/2560

Rice Disease and Insect pest Identifition Algorithm with Shuffle Attention Mechanism Reviewed International journal

Yuliang Gao, Lifeng Zhang, Li Bin

Proceedings of the 8th IIAE International Conference on Intelligent Systems and Image Processings 2021 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 159 - 164 2021.09

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kobe 2021.09.06 - 2021.09.10

This paper presents an identififition algorithm of rice diseases and insect pests based on convolutional neural network. Aiming at the problems of large amount ofparameters and low accuracy of traditional algorithm. The traditional convolution neural network Inception-ResNetV2 and the shuffle attention module are combined. It achievesthe effects of greatly reducing parameters and obviously improving accuracy. Compared with traditional models ResNet50 and Inception-ResNetV2, the highest accuracyreaches 95.24%, the parameters in the self-made data set decreased by 47.6% and 34.02%, and the accuracy increased by 1.49% and 0.9%, respectively.

DOI： 10.12792/icisip2021.028

Other Link： https://www2.ia-engineers.org/conference/index.php/icisip/icisip2021/paper/view/2480

Self-supervised monocular depth estimation with occlusion mask and edge awareness Reviewed International journal

Shi Zhou, Miaomiao Zhu, Zhen Li, He Li, Mitsunori Mizumachi, Lifeng Zhang

Artificial Life and Robotics ( Springer ) 26 ( 3 ) 354 - 359 2021.05

Authorship：Last author Language：English Publishing type：Research paper (scientific journal)

Depth estimation is one of the basic and important tasks in 3D vision. Recently, many works have been done in self-supervised depth estimation based on geometric consistency between frames. However, these research works still have difficulties in ill-posed areas, such as occlusion areas and texture-less areas. This work proposed a novel self-supervised monocular depth estimation method based on occlusion mask and edge awareness to overcome these difficulties. The occlusion mask divides the image into two classes, making the training of the network more reasonable. The edge awareness loss function is designed based on the edge obtained by the traditional method, so that the method has strong robustness to various lighting conditions. Furthermore, we evaluated the proposed method on the KITTI datasets. The occlusion mask and edge awareness are both beneficial to find corresponding points in ill-posed areas.

DOI： 10.1007/s10015-021-00685-z

Kyutacar

Scopus

Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85106505382&origin=inward

Proposal of Cursor Operation Method by Facial Movements for the Handicapped Reviewed

Shunya Mito, Lifeng Zhang

Proceedings of the 9th IIAE International Conference on Industrial Applications Engineers 2021 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 3 - 7 2021.03

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kitakyushu 2021.03.26 - 2020.03.30

In recent years, the number of people with physical disabilities has been on the rise, with about half of them being physically disabled limb. A physically disabled person is a person with a physical disability that interferes with daily life, such as missing or malfunctioning fingers. This study focused on the use of computers in daily life. A person with a limb disability used a PC.Earlier studies that made "eye contact", "voice input", "facial expressions using a 3D camera". Those problems are the need to keep the distance between the face and the camera constant, the inability to protect privacy when people are around, and that 3D cameras are difficult and expensive to install. In this study, by using a webcam as input and machine learning, image processing techniques. Developing a system that allows a handicapped person to operate a computer with only facial movements without the use of arms or hands. We also evaluated the operability, improving the problems of previous studies.

DOI： 10.12792/iciae2021.003

Other Link： https://www2.ia-engineers.org/conference/index.php/iciae/iciae2021/paper/view/2396

Speech Sentiment Analysis Based on Basic Characteristics of Speech Signal Reviewed

Zijun Yang, Lifeng Zhang, Seiichi Serikawa

Proceedings of the 9th IIAE International Conference on Industrial Applications Engineers 2021 ( The Institute of Industrial Applications Engineers ) 1 ( 1 ) 259 - 263 2021.03

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kitakyushu 2021.03.26 - 2020.03.30

As the pace of life continues to accelerate, people's life pressure is increasing. With the accumulation of time, people's mental and psychological conditions have been affected to a certain extent. These mental illnesses will not cause any impact under normal circumstances, but once they break out, they will cause trauma that cannot be ignored in life or even in society. Therefore, we hope to design a system program that can chat with humans in daily life, and it can feel the human's emotional changes in daily conversations. When humans have negative emotions, it can comfort us in time and even warn humans when our negative emotions reach a certain limit. When humans have positive emotions, it can give humans affirmative approval and encouragement. Based on this concept, we must first analyze the different emotions that humans design in daily conversations. This article is mainly based on the basic characteristics of audio signals to judge the user's emotional changes. The database we use is six different emotional voices recorded by four voice actors, and each voice contains 50 single sentences for emotional recognition analysis.

DOI： 10.12792/iciae2021.047

Other Link： https://www2.ia-engineers.org/conference/index.php/iciae/iciae2021/paper/view/2445

　 PREV - NEXT 　

<ZHANG Lifeng>

Papers - ZHANG Lifeng