Object Recognition Algorithm Research Articles

A Deep Learning-Enhanced Multi-Modal Sensing Platform for Robust Human Object Detection and Tracking in Challenging Environments

In modern security situations, tracking multiple human objects in real-time within challenging urban environments is a critical capability for enhancing situational awareness, minimizing response time, and increasing overall operational effectiveness. Tracking multiple entities enables informed decision-making, risk mitigation, and the safeguarding of civil-military operations to ensure safety and mission success. This paper presents a multi-modal electro-optical/infrared (EO/IR) and radio frequency (RF) fused sensing (MEIRFS) platform for real-time human object detection, recognition, classification, and tracking in challenging environments. By utilizing different sensors in a complementary manner, the robustness of the sensing system is enhanced, enabling reliable detection and recognition results across various situations. Specifically designed radar tags and thermal tags can be used to discriminate between friendly and non-friendly objects. The system incorporates deep learning-based image fusion and human object recognition and tracking (HORT) algorithms to ensure accurate situation assessment. After integrating into an all-terrain robot, multiple ground tests were conducted to verify the consistency of the HORT in various environments. The MEIRFS sensor system has been designed to meet the Size, Weight, Power, and Cost (SWaP-C) requirements for installation on autonomous ground and aerial vehicles.

Solution of Real-Time Traffic Signs Detection Problem for Autonomous Vehicles by Using YOLOV4 And Haarcascade Algorithms

Unmanned systems are increasingly used today to facilitate our daily lives and use time more efficiently. Therefore, this rapidly emerging and growing technology appears in every aspect of our lives with its various functions. Object recognition algorithms are one of the most important functions that we often encounter in these systems. Autonomous vehicle technologies are the latest and fastest growing technology among unmanned systems. In this study, we investigate the success rates of two different algorithms for recognizing traffic signs and markings that can be used for partially or fully autonomous vehicles. In this study, two different solutions to the problem of recognizing the signs for fully autonomous and fully autonomous vehicles, respectively, were presented and the correct identification of the markers was evaluated. The work was performed in real-time. Two different concepts were used for these products. An enclosed space where an ideal lighting environment is provided for the evaluation of models should be visualized. In addition, for the general recognition of the models, the test procedures were performed with a dataset obtained from the users and it was computed for the general recognition. In addition, this study aims to provide a better understanding of the basic working principles, the differences between machine learning and deep learning, and the contents of object recognition processes.

Real-time Fast Selection System with Object Recognition and TSP algorithms

The stage before the conversion of agricultural products into post-harvest consumer products is the process of separating the raw products into appropriate classes. Today, this difficult manual separating process is a process in which a large number of workers work at an intense pace on the product line and the workforce is intensively spent. Disruptions in separating as a result of carelessness cause product loss, loss of time and cost increases. In this study, as an alternative to manual separating processes, a real-time separating system, which detects the products in the factory band with object recognition methods and enables fast positioning of the separating tool on the products, works simultaneously with object recognition and traveling salesman problem algorithms has been created. In this way, a low-budget separating system is recommended for large selecting processes with a time- and cost-effective selecting model. In the study, the creation of a real-time fast separating system with the support of the traveling salesman algorithm, performance evaluation and research and findings on the fast separating model are presented.

Towards Recognition of Human Actions in Collaborative Tasks with Robots: Extending Action Recognition with Tool Recognition Methods.

This paper presents a novel method for online tool recognition in manual assembly processes. The goal was to develop and implement a method that can be integrated with existing Human Action Recognition (HAR) methods in collaborative tasks. We examined the state-of-the-art for progress detection in manual assembly via HAR-based methods, as well as visual tool-recognition approaches. A novel online tool-recognition pipeline for handheld tools is introduced, utilizing a two-stage approach. First, a Region Of Interest (ROI) was extracted by determining the wrist position using skeletal data. Afterward, this ROI was cropped, and the tool located within this ROI was classified. This pipeline enabled several algorithms for object recognition and demonstrated the generalizability of our approach. An extensive training dataset for tool-recognition purposes is presented, which was evaluated with two image-classification approaches. An offline pipeline evaluation was performed with twelve tool classes. Additionally, various online tests were conducted covering different aspects of this vision application, such as two assembly scenarios, unknown instances of known classes, as well as challenging backgrounds. The introduced pipeline was competitive with other approaches regarding prediction accuracy, robustness, diversity, extendability/flexibility, and online capability.

A novel recognition of artificial intelligence in regional anaesthesia

Mechanised analysis of algorithms for problem solving, object recognition and decision-making assistance defines the general concept underlining artificial intelligence (AI).[1] The former presents notable applications in perioperative medicine across diverse situations like depth of anaesthesia monitoring, closed-loop anaesthesia delivery, risk-predictive models, pain-management through objective analgo-scoring, operating-room planning logistics and, guidance of perioperative ultrasonography. The recent Food and Drug Administration (FDA) approval[2] of the Scan Nav™ Anatomy Peripheral Nerve Block (PNB) system (Intelligent Ultrasound Group, Cardiff, United Kingdom)[3] only emerges as a novel recognition of the potential of AI in regional anaesthesia. Worthwhile to mention, Hemmerling et al.[4] described an encouraging seminal use of the robotic ultrasound-guided nerve block system or the Magellan system, almost a decade ago. Employing the same system, Morse et al.[5] went on to compare the performance parameters (learning curve and success rates) and subjective performance variability of manual versus robot-assisted ultrasonographic nerve blocks in a phantom model. The authors found that assistance of nerve blocks, enabled significant faster needle guidance (1.8 [1.6] seconds vs. 0.3 [0.3] seconds; P = 0.007), than for manual blocks. Inter-subject performance variability was also found to be considerably attenuated. Specific to the context of the recently FDA approved software medical device, ScanNav™ Anatomy PNB system uses deep-learning AI technology. During performance of a peripheral nerve block the ScanNav™ system identifies and grades a live ultrasound image using this technology. A distinctive colour overlay of the key sono-anatomical structures during a live ultrasound scan is a characteristic valuable attribute of the system.[3] This enhances standardisation and accuracy of the ultrasound image interpretation. This device colour highlights anatomy of 10 vital regions relevant to ultrasound-guided regional anaesthesia namely Interscalene, Supraclavicular, Axillary and Superior Trunk Brachial plexus of the upper limb; Erector Spinae Plane, Rectus Sheath and Supra-inguinal Fascia Iliaca sheath of truncal structures; and Femoral Nerve, Adductor canal and Popliteal nerve sheath of the lower limb. Furthermore, a 3-dimension inbuilt reference material provides a review of the relevant anatomy, probe-positioning technique; and demonstrates a virtual scan of the selected anatomical region. The system enables adjustments in the intensity of coloured highlighting according to user preference and allows for viewing the unmodified and highlighted images side-by-side for better comprehension of the sono-anatomy. The use of ScanNav™ system thereby facilitates as a labelling device alongside real-time anatomy visualisation during regional anaesthesia. The system also has inbuilt virtual scans that assist the clinician functioning as a 3D reference material of the relevant PNB anatomy scanned. Notably so, the software is compatible with the general-purpose ultrasound systems in addition to being a stand-alone device. Talking of the literature relevant to the ScanNav™ Anatomy PNB system, Bowness et al.[6] compared 240 ultrasound scans performed across 15 expert versus non-expert regional anaesthesia providers on evaluating its' role in promoting the uptake and generalisability of ultrasound-guided regional anaesthesia. The authors outlined that non-experts were more likely to submit positive feedback and less likely to submit negative reviews than the experts (P = 0.001). Positive feedbacks were obtained on the role of the device in training (37/60, 61.7%) and utility for teaching (30/60, 50.0%). Meanwhile the non-experts appreciated the anatomic structure identification and learning, as the contextual benefits, some nonexperts and experts thought the device could decrease the scanning confidence. Additionally, expert providers reported a potentially increased, yet insignificant, risk in 12/254 (4.7%) vs. 8/254 (3.1%, P = 0.362) scans for needle trauma to “safety” structures. The discussion elaborates the tremendous opportunity AI has to offer in guiding regional anaesthesia, awaiting prospective validation outside the core 10 block areas. The utility of this device can be further extended to sympathetic blocks like the stellate-ganglion, celiac-plexus and other blocks for treatment of chronic pain syndromes like complex regional pain syndrome. The times might not be distant when perioperative physicians have routine access to ultrasound machines integrated with AI-assisted colour reframing for anatomical landmark “coding” side by side with the present B-mode, M-mode and colour/pulse Doppler profiling.

Object Detection with Voice Guidance to Assist Visually Impaired Using Yolov7

Abstract: Object recognition technology has revolutionized various domains such as autonomous vehicles, industrial facilities, and many more. However, the visually impaired individuals, who are most in need of this technology, have not been able to utilize it effectively. Therefore, this paper presents a deep learning-based object detection system that is specifically designed for the blind community. The system incorporates the YOLOV7 (You Only Look Once) algorithm for object recognition and uses text-to-speech (TTS) technology to provide a voice-guidance technique that conveys information about the objects around them. The main objective of the proposed system is to empower visually challenged individuals to independently identify objects in a particular space without relying on external assistance. The system's efficiency and performance are thoroughly scrutinized through experiments to validate its accuracy and effectiveness. The proposed system employs technical terms such as deep learning, object recognition, YOLOV7 algorithm, and also image classification techniques for feature identification and extraction of the video frames and categorize them into the respective classes. It uses COCO dataset, which consist of around 123,287 hand-labeled images classified into 80 categories, has been utilized to train the proposed system. In conclusion, the suggested system for object detection uses cutting-edge deep learning techniques along with voice guidance methods to assist individuals who are visually impaired in identifying objects independently. The performance of the system is evaluated through experiments, demonstrating its effectiveness and potential for real-world applications

RETRACTED ARTICLE: Evaluation of image segmentation and multi class object recognition algorithm based on machine learning

This paper mainly updates the image segmentation and multi-target recognition algorithm through the support vector machine algorithm in ML, and verifies the reliability of the new proposed algorithm through simulation experiments, and determines that the ML-based image segmentation and multi-target recognition algorithm can have stronger processing capability in image recognition and processing. Finally, the performance of traditional image segmentation and multi class object recognition algorithms and the image segmentation and multi class object recognition algorithms proposed in this paper were compared through experiments. The results showed that the performance of the ML based image segmentation and multi class object recognition algorithms proposed in this paper has increased by 29.1% on average in all aspects.

SO-YOLOv5: Small object recognition algorithm for sea cucumber in complex seabed environment

Underwater object recognition is a prerequisite for the realization of automated seafood harvesting. To solve the problems of low recognition accuracy and poor generalization ability of existing underwater small object, this paper takes sea cucumber as research objects to research the small object recognition approach in complex underwater environment. In this paper, a sea cucumber dataset is established to solve the problem of lacking sea cucumber dataset in the real seabed environment. A deep learning model SO-YOLOv5 is proposed based on YOLOv5 for underwater small object recognition, which improves the recognition ability of the algorithm for different sizes of objects in complex environment. The proposed model embeds a large-scale feature extraction layer in structure to increase the detection ability of small objects by referring to the idea of the Bi-directional Feature Pyramid Network. SO-YOLOv5 fuses the features between deep and shallow layers and balances the feature information of different scales, and integrates the Coordinate Attention mechanism to enhance the sensitivity of the algorithm to the direction and position information. The experimental results illustrate that the mAP of the proposed approach achieves 95.47% for sea cucumber recognition in the complex seabed environment, which is 3.42%, 6.79%, and 5.46% higher than the traditional Faster R-CNN, SSD, and YOLOv5 algorithms, respectively. This research not only provides an effective approach for sea cucumber recognition in complex environment but also has certain reference significance for the recognition of small objects in other complex environments. In addition, the proposed approach has practical application value to improve intelligence level in aquaculture.

Bicycle Riding Side Identification by Image Sensing of Asymmetric-Shaped Road Marking Orientation

This paper describes a bicycle riding side identification method by image sensing of the asymmetric-shaped road marking orientation in order to realize an advanced bicycle navigation system and a bicycle riding safety support system. Firstly, we define the requirements of road markings in this method as existing in bicycle riding space, having a vertical asymmetrical shape, and being available to capture by a bicycle-mounted camera. Secondly, we construct a prototype system using an object detection and recognition algorithm to confirm the feasibility of this method for identifying which side of the roadway the bicycle rides on. Then, the experiment to test feasibility is conducted using the images of arrow feather road markings photographed in the real environment. As a result, it is confirmed that the bicycle riding side could be identified with high accuracy. Consequently, we show the feasibility of the bicycle riding side identification method, which does not use dedicated sensors considering the flexibility of bicycle mobility, and by using asymmetric-shaped road marking orientation and a common camera.

The Deep Learning based Smart Navigational Stick for Blind People

Blind and visually impaired people find difficulty in detecting obstacles and recognizing people in their way, which makes it dangerous for them to walk, to work, or to go in a crowded area/place. They have to be cautious all the time to move, while avoiding any solid obstacles in their way. Typically, they use different aid devices to reach their destination or to accomplish their daily task. The normal stick is useless for blind and visually impaired people since it cannot detect barriers or people's faces. Visually impaired individuals are unable to distinguish between different types of objects in front of them. They are unable to gauge the size of an object or its distance from them. Several works have been done by public individuals and scientific investigators but their work is dearth in technological aspect. This technological aspect need to be addressed by adding artificial intelligence (AI). This prototype aims to help blind and visually impaired individuals in several aspects to simply obtain/perform everyday tasks and help these individuals to live with the same confidence as sighted people live.Therefore, this study inclined deep learning Mobile-Net Single Shot MultiBox detection (SSD) algorithm for object recognition and Dlib library for face recognition. Subsequently, the proposed solution is using an Open CV and Python. Additionally, Ultrasonic sensors are used for distance measurement, which can be a great help for visually impaired people. These components are grouped together to work effectively and efficiently for the development of visually impaired people. The recognition procedure was revealed through headphones, which notifies the visually impaired when face or any object get recognized. Inclusively, the innovative solution would be a great aid for the blind and visually impaired individuals. As a result, to test and validate the accuracy of the smart navigational stick, several experiments have been conducted on a range of objects and faces. Hence, this study’s modified navigational system was adequate and valid for visually impaired people.

Deep Learning for Highly Accurate Hand Recognition Based on Yolov7 Model

Hand detection is a key step in the pre-processing stage of many computer vision tasks because human hands are involved in the activity. Some examples of such tasks are hand posture estimation, hand gesture recognition, human activity analysis, and other tasks such as these. Human hands have a wide range of motion and change their appearance in a lot of different ways. This makes it hard to identify some hands in a crowded place, and some hands can move in a lot of different ways. In this investigation, we provide a concise analysis of CNN-based object recognition algorithms, more specifically, the Yolov7 and Yolov7x models with 100 and 200 epochs. This study explores a vast array of object detectors, some of which are used to locate hand recognition applications. Further, we train and test our proposed method on the Oxford Hand Dataset with the Yolov7 and Yolov7x models. Important statistics, such as the quantity of GFLOPS, the mean average precision (mAP), and the detection time, are tracked and monitored via performance metrics. The results of our research indicate that Yolov7x with 200 epochs during the training stage is the most stable approach when compared to other methods. It achieved 84.7% precision, 79.9% recall, and 86.1% mAP when it was being trained. In addition, Yolov7x accomplished the highest possible average mAP score, which was 86.3%, during the testing stage.

Sorokh-Poth: A Balanced Bangladeshi Road Vehicle Image Dataset Integrated With a Detection System Based On Deep Convolutional Neural Networks

When it comes to the administration of highways, the identification of intelligent vehicles is becoming an increasingly important component. Applications for vehicle detection are essential for both military and civilian usage, including the monitoring of highway traffic, administration, and the control of urban traffic. The vast amount of vehicle image data available online has the potential to encourage the development of increasingly sophisticated object recognition and classification models and algorithms. But finding an organized, balanced, and meaningful dataset continues to be a major challenge. This study proposes Sorokh-Poth, a new complete balanced image dataset based on Bangladeshi road travel that is compatible with a number of CNN-based architectures. The majority of the photographs in the collection were taken with a smartphone. The dataset includes 9,809 classified and annotated images of ten different types of vehicles, including autorickshaws, bicycles, buses, cars, CNG-powered vehicles, lagoon rickshaws, trucks, and vans. This research work utilizes the residual network ResNet-50 model, a CNN-based architecture. Here, features specific to the type of vehicle were automatically retrieved and grouped. Accuracy, precision, recall, and f1-score were just a few of the metrics used for evaluation during the research. The proposed model exhibited an increasing accuracy despite the vehicles shifting physical characteristics. The purpose of this work is to implement various CNN-based residual networks ResNet-50 and ResNet-150V2. Our proposed ResNet-50 model achieves an accuracy of 98.00% in the detection of native Bangladeshi vehicles, according to result comparisons and evaluations.

Performance Evaluation of MEMS-Based Automotive LiDAR Sensor and Its Simulation Model as per ASTM E3125-17 Standard

Measurement performance evaluation of real and virtual automotive light detection and ranging (LiDAR) sensors is an active area of research. However, no commonly accepted automotive standards, metrics, or criteria exist to evaluate their measurement performance. ASTM International released the ASTM E3125-17 standard for the operational performance evaluation of 3D imaging systems commonly referred to as terrestrial laser scanners (TLS). This standard defines the specifications and static test procedures to evaluate the 3D imaging and point-to-point distance measurement performance of TLS. In this work, we have assessed the 3D imaging and point-to-point distance estimation performance of a commercial micro-electro-mechanical system (MEMS)-based automotive LiDAR sensor and its simulation model according to the test procedures defined in this standard. The static tests were performed in a laboratory environment. In addition, a subset of static tests was also performed at the proving ground in natural environmental conditions to determine the 3D imaging and point-to-point distance measurement performance of the real LiDAR sensor. In addition, real scenarios and environmental conditions were replicated in the virtual environment of a commercial software to verify the LiDAR model's working performance. The evaluation results show that the LiDAR sensor and its simulation model under analysis pass all the tests specified in the ASTM E3125-17 standard. This standard helps to understand whether sensor measurement errors are due to internal or external influences. We have also shown that the 3D imaging and point-to-point distance estimation performance of LiDAR sensors significantly impacts the working performance of the object recognition algorithm. That is why this standard can be beneficial in validating automotive real and virtual LiDAR sensors, at least in the early stage of development. Furthermore, the simulation and real measurements show good agreement on the point cloud and object recognition levels.

Spectrogram-based classification on vehicles with modified loud exhausts via convolutional neural networks

Hong Kong is one of the densest cities in the world with busy roads close to residential areas. Excessive noise emitted by vehicles with modified loud exhausts causes further noise annoyance during late night. Existing technologies are inefficient for detecting these undesirable vehicles due to various constrains such as the use of sound level meter requires manual spectrum analysis, and acoustic camera is expensive, bulky for low frequency detection and computationally heavy. In this research, the noise spectral characteristics at the peak sound pressure levels from vehicles observed with and without modified loud exhausts passing-by the measurement points were collected and scientifically analyzed. A spectrogram-based sound recognition prototype with low-cost design was therefore developed in assisting the detection of vehicles with modified loud exhausts. A large amount of audio data was converted into spectrograms by short-time Fourier transform for machine learning and a trained convolutional neural network model based on AlexNet architecture utilized spectrograms as the inputs for classifying whether the pass-by vehicles are with modified loud exhausts. The prototype also deploys YOLO v4 Tiny algorithm for object recognition and is able to distinguish the types of pass-by vehicles i.e. passenger car (which is the target for detection), goods vehicle, and motorcycle for enhancing the usability. The results show that the prototype had reached an overall 96% accuracy in the detection of passenger cars with and without modified loud exhausts.

Performance analysis of object detection algorithms for robotic welding applications in planar environment

ABSTRACT Weld seam detection and tracking using intelligent techniques is a common task in robotic welding. In this study, vision-based robotic welding is implemented using a 2D Camera to reduce the experimental cost in a planar environment. Firstly, different multi-class object recognition algorithms namely Faster Region-Based Convolutional Neural Network (Faster R-CNN), Single Shot Detector (SSD) and You Look Only Once (YOLOv3) are trained with 1530 manually labeled images. To recognize the weld seam under different brightness levels addressing arc light and splash, Contrast-Limited Adaptive Histogram Equalization (CLAHE) method has been used. Further, an extensive experimental study is conducted to find the optimal hyperparameters among three deep learning models. The performance metrics like Accuracy, Root Mean Square Error, Precision, Recall, Mean Average Precision and F1-score are calculated for different test cases. Finally, homogeneous coordinate transformation is implemented to obtain robot coordinates from weld seam pixel coordinates. The analysis is also extended to detect the bad welding conditions for improper shapes and track the weld seam for good welding shapes using TAL BRABO manipulator. It was found that the weld shapes were accurately detected and tracked precisely with 99.9% accuracy using YOLOv3 than Faster R-CNN and SSD.

Remote Big Data Management and Visual Imagery Tools, Multisensor Fusion and Dynamic Routing Technologies, and 3D Space Mapping and Object Recognition Algorithms on Blockchain-based Metaverse Platforms

Remote Big Data Management and Visual Imagery Tools, Multisensor Fusion and Dynamic Routing Technologies, and 3D Space Mapping and Object Recognition Algorithms on Blockchain-based Metaverse Platforms

A Multi-Purpose Realistic Haze Benchmark With Quantifiable Haze Levels and Ground Truth.

Imagery collected from outdoor visual environments is often degraded due to the presence of dense smoke or haze. A key challenge for research in scene understanding in these degraded visual environments (DVE) is the lack of representative benchmark datasets. These datasets are required to evaluate state-of-the-art object recognition and other computer vision algorithms in degraded settings. In this paper, we address some of these limitations by introducing the first realistic haze image benchmark, from both aerial and ground view, with paired haze-free images, and in-situ haze density measurements. This dataset was produced in a controlled environment with professional smoke generating machines that covered the entire scene, and consists of images captured from the perspective of both an unmanned aerial vehicle (UAV) and an unmanned ground vehicle (UGV). We also evaluate a set of representative state-of-the-art dehazing approaches as well as object detectors on the dataset. The full dataset presented in this paper, including the ground truth object classification bounding boxes and haze density measurements, is provided for the community to evaluate their algorithms at: https://a2i2-archangel.vision. A subset of this dataset has been used for the "Object Detection in Haze" Track of CVPR UG2 2022 challenge at https://cvpr2022.ug2challenge.org/track1.html.

Bronchop Neumonia Detection Using Novel Multilevel Deep Neural Network Schema

Pneumonia is a dangerous disease that can occur in one or both lungs and is usually caused by a virus, fungus or bacteria. Respiratory syncytial virus (RSV) is the most common cause of pneumonia in children. With the development of pneumonia, it can be divided into four stages: congestion, red liver, gray liver and regression. In our work, we employ the most powerful tools and techniques such as VGG16, an object recognition and classification algorithm that can classify 1000 images in 1000 different groups with 92.7% accuracy. It is one of the popular algorithms designed for image classification and simple to use by means of transfer learning. Transfer learning (TL) is a technique in deep learning that spotlight on pre-learning the neural network and storing the knowledge gained while solving a problem and applying it to new and different information. In our work, the information gained by learning about 1000 different groups on Image Net can be used and strive to identify diseases.

Behavior of Delivery Robot in Human-Robot Collaborative Spaces During Navigation

Navigation is an essential skill for robots. It becomes a cumbersome task for the robot in a human-populated environment, and Industry 5.0 is an emerging trend that focuses on the interaction between humans and robots. Robot behavior in a social setting is the key to human acceptance while ensuring human comfort and safety. With the advancement in robotics technology, the true use cases of robots in the tourism and hospitality industry are expanding in number. There are very few experimental studies focusing on how people perceive the navigation behavior of a delivery robot. A robotic platform named “PI” has been designed, which incorporates proximity and vision sensors. The robot utilizes a real-time object recognition algorithm based on the You Only Look Once (YOLO) algorithm to detect objects and humans during navigation. This study is aimed towards evaluating human experience, for which we conducted a study among 36 participants to explore the perceived social presence, role, and perception of a delivery robot exhibiting different behavior conditions while navigating in a hotel corridor. The participants’ responses were collected and compared for different behavior conditions demonstrated by the robot and results show that humans prefer an assistant role of a robot enabled with audio and visual aids exhibiting social behavior. Further, this study can be useful for developers to gain insight into the expected behavior of a delivery robot.

Использование технического зрения в гибких производственных системах для определения координат беспорядочно расположенных объектов

The article is devoted to the development of an object recognition algorithm using technical vision using Python and the OpenCV computer vision library. The article presents a program that allows you to set the coordinates of an object arbitrarily located in the field of view of the camera, as well as determine its orientation. This data will allow you to perform an effective capture of the object by the grip of the manipulator. In modern mechanical engineering, tasks of this kind are quite relevant, they make it possible to increase the autonomy of flexible production systems and make production safer.