Search

Pothole detection is a challenging task as it involves inhomogeneity in intensity, complex topologies, contrast less, and noise rich background. Still, it's a very important task concerning the maintenance of roads. Potholes are being detected using computer vision which generally involves analysis of a two-dimensional image of the road or modeling of the road surface. Provided both these methods are independently used. The accuracy of this method of pothole detection is far from the top point. These bring the usage of digital image processing which develops the detection of potholes with at most accuracy. Digital Image processing provides strong receptivity, usage of huge actions, and broad benefits. In this paper, we propose a system that supports digital image processing techniques of high efficiency to result in consistent and accurate detection of potholes. This project studies an intelligent digital industrial pothole detection mechanism to increase human safety, detection capability and decrease the factor of risk. Firstly, the sources of images from the road are pre-processed to remove noise and other irrelevant features. Then a training model was developed to train the system to classify between path holes and normal cracks using convolution neural networks and MATLAB. Also, an IoT device is used in this system which updates the detected potholes on the road to the respective web server along with the location coordinates, as a part to notify a complaint to check. This system works on providing a better pothole detection technique and also updating the same to the concerned. [ABSTRACT FROM AUTHOR]

Various object detection models are presented in this paper. Definition of object is well defined in order to understand object detection models. Now-a-days the object detection which is computer vision and image processing related technology is widely used. Applications of object detections are infinite as it can be used all over like objects tracking, surveillance of videos, face recognition, detection of anomaly, counting of people, and the records still continue to many more. This paper compares various object detection algorithms the basic splitting is based on machine learning based and deep learning based. Fundamental objective of the paper focuseson Single stage object detectors (SSD) like You Only Look Once (YOLO) and Two stage object detectors as RCNN, Fast RCNN, etc. This paper's main goal is to highlight such object detection models and their use in various fields. [ABSTRACT FROM AUTHOR]

One of the areas of computer vision is superpixel segmentation, and the widespread use of superpixels in a wide range of practical applications shows that it has unique properties. A superpixel is created by combining pixels with similar content, which significant reduce the computational complexity of the algorithm at the next stages of image processing. After the superpixel is built, it is possible to solve problems such as classification, reduction of image dimensions, construction of spectral combination, selection of a certain channel, noise reduction, and change detection based on the application of superpixel to spectral images. This paper presents an analysis of SLIC, SEEDS, LSC, DFIC, AHP, DRW, WSS, BACA, and superpixel segmentation approaches on remote sensing images. The analysis of approaches was considered according to the characteristics of the color model, scale, regularity, connection, smoothness, texture, and boundary strength. Approaches were evaluated using the criteria of Undersegmentation error, Achievable segmentation accuracy, and Boundary Recall. The study revealed that superpixel segmentation techniques that employ all of the spectral image channels have not received enough attention; instead, they are typically applied using PCA to shrink the size or convert them to color models. Additionally, it is important to investigate the issue of applying deep learning techniques to the implementation of superpixel segmentation in remotely sensing spectral images. [ABSTRACT FROM AUTHOR]

The detection of image edges is a major challenge in image processing and computer vision, since it needs to dispense of unnecessary pixels while preserving significant ones which indicate the image's structure. An optimized algorithm for image edge detection is introduced in this study, which uses a Genetic Optimization (GO) Algorithm to seek out sources of the maximum pixel value to optimally allocate the image edges. The algorithm simulates the evaluation of various values and consists of three operations: selection, crossover operations, and thresholding the optimized value. The community is made up of these evaluations which are conducted at each stage to select the fittest individuals and eliminate the less fit ones. By repeating this process, GO eventually identifies the image edges and eliminates non-edge pixels. The results of the study show that the proposed algorithm successfully addresses the challenges associated with image edge detection. The limitations of the research are also discussed, and potential areas for future research are suggested. This algorithm is expected to be useful for solving complex problems. [ABSTRACT FROM AUTHOR]

The automatic safety system for the swimming pool system has been proposed using Computer vision and IoT. So many drowning incidents are happening, in order to avoid this drowning; the system tracks the position of the person in the water using videos. In this work, we have tracked the particular person in the water and if the person is drowning the system will give the alert message. Image processing algorithms were used to classify the drowning objects or persons. IoT and transfer learning has been combined to identify the pool safety automatically. Transfer learning model uses the pretrained model "ImageNet" that has more than 14,000 images for classification. Using ImageNet, classification of human or objects drowning can be performed correctly. The classification results have been sent to the mobile device if required. [ABSTRACT FROM AUTHOR]

In the face of the upcoming Corona pandemic in 2020, which continues to this day, adequate protection for all citizens has been very important and continues to be essential. An easy way to minimize the risk of infection in public, but poorly ventilated places, are FFP masks. The basic material from which FFP masks are made is nonwoven. The quality of the produced nonwoven is the basis for the correct function and thus the protection against infection. The present research work investigates methods of computer vision to detect contaminations and damages on the nonwoven. Both supervised and semi supervised methods are evaluated. For this work, an inspection system consisting of two separate acquisition systems was developed for image data acquisition. The first system is suitable for the visible wavelength and has a theoretical resolution of 2.4 µm per pixel. The second camera system is designed for the near infrared range and has a theoretical resolution of 5.5 µm per pixel. The acquisition system collects an image data set comprising 1,760 images with 920 images of defect-free nonwoven samples and 840 images of defective nonwoven samples. First, the wavelength range suitable for optical inspection of uncoated nonwoven samples is investigated. A further investigation dealt with the question of whether the reflected light or transmitted light method is more suitable for optical inspection. Finally, coated nonwoven is also inspected using the reflected light method. An investigation using the transmitted light method is not possible, as the material is almost non-transparent. Despite the small amount of data, very good results were achieved. Machine learning methods from the field of image processing are usually classified as deep learning. This means that the large network architectures require very large amounts of data in order to learn complex patterns. Publicly available datasets for method evaluation typically consist of over 1,000 images per class or defect. The nonwoven samples provided in this work and the resulting image database is about a factor of six smaller than is actually intended for the methods used. In this work it has been possible to achieve inspection accuracies of 97.5 %. [ABSTRACT FROM AUTHOR]

The Corona virus wreaked havoc on the worldwide health system. The transmission of Corona virus can be reduced and prevented by wearing a face mask. To eradicate the spreading of Corona virus among the individuals we proposed a system that can detect the individual who are not wearing mask and alert the individual and we employed a dataset with and without masks photos from the public domain Kaggle facial mask detection dataset, and a predefined model MobileNetV2 and CNN Algorithm are accomplished for real-time face detection. We use Python, OpenCV, Tensor Flow, and Keras to develop a facemask detector in depth using computer vision techniques. Our objective is to apply image processing techniques and deep learning algorithms to analyse an image or video stream to ascertain whether an individual is wearing a face mask or not. If the person is not, an alarm warning will be generated. In proposed model Swish activation function applied in place of ReLu to generate an optimal accuracy of 99.81 [ABSTRACT FROM AUTHOR]

Drones are being utilized for a variety of functions, including military operations, agriculture, aerial photography, surveillance, and remote sensing. The widespread misuse of drones or UAVs, results in security, technological and public safety concerns that must be addressed and controlled. Security organizations are always looking for technology and sophisticated systems that can identify drones and secure the region around them. The present study deals a drone detection technique which is developed using Convolution Neural Networks. YOLOv4 and a convolution neural network were used to train the model for object detection. A small rocket-assisted missile can be used to capture the detected drone. Two processing units are in charge of the proposed missile. The first processing unit will employ image processing in real-time, while the remainder will be handle by the second processing unit, including missile control, monitoring, and targeting. Computer vision algorithms and an electro-optical sensor (a camera) allow the missile to track the target after it leaves the missile launcher. Tracking has been done to determine the precise location and speed of the target. It can be carried out by changing where the missile is and where it is going after it leaves the launcher. [ABSTRACT FROM AUTHOR]

Image restoration is the process in which a corrupted image is restored to its original quality. Since there are plenty of computer vision applications where images are widely used and this is very important to restore noisy or corrupted images for high visual quality and accuracy of image processing-based applications. We aim at providing multiple techniques that caters to the needs of images of varied quality. The techniques to be implemented in this project for image restoration include different deep learning approaches such as SRCNN, SRGAN, deep learning attributes, Ffmpeg (Fast Forward Moving Pictures Experts Group), and bilateral filter. A unified method is given for various types of inputs as it can repair the damaged images and enhance. Super resolution, dehazing, colorization and inpainting are the blocks through which the input go through and get an escalated output. It can support grayscale and colored images. The application is built using Python data science platform. Enables users to give any image as input and output can be seen on the same screen. For human eye perception, sometimes, output and input may appear same. In order to show the quality improvement, we use statistical measures such as PSNR (Peak signal to noise ratio) and SSIM (Structural similarity index measure). These metrics show the performance improvement with quantitative results for proof of the concept. This project also has provision to enhance short videos clips. As the main goal is refinement. [ABSTRACT FROM AUTHOR]

Currently, many companies are competing to produce robots that can help humans. Robots can work with high accuracy, efficiency, and low power in the production of goods. Automated Guided Vehicle is one form of robot that is useful in industry. Automated Cars are another form of functional industrial robot that transports with programmable paths and directions automatically. In this paper, we will discuss the navigation system in Automatic Cars based on previously mapped paths. The robot wheel serves to adjust the direction and position of the robot. The automatic movement can be carried out using computer vision. Warehouses that are not neatly arranged cause the navigation system to be less than optimal. The types of controllers used include classical controllers, fuzzy logic, artificial neural networks, and H∞. The purpose of this research is to produce a precise car front view monitoring system using image processing with high enough quality for automatic car navigation. The benefit for science is the development of automatic cars, which so far have not been widely carried out in Indonesia. The results of this study can be useful for the development of methods to build an automatic car system. The benefit for the electronics industry is the availability of control equipment designs to be able to control cars automatically. The effect on society is in the form of future vehicles that can move automatically without a driver. Based on the computer vision system test for Raspberry Pi-based car navigation, it can be ascertained that the car works well and has a small physique, so it looks neat and concise. The movement of the car from the first coordinate point to the next coordinate point is very precise and precise. By installing the L298N motor driver module, it is easier to set the program in python, so that the results sent to the DC motor can be used directly. Using a DC motor that has an excellent performance in terms of torque and speed. [ABSTRACT FROM AUTHOR]

Image fusion is the integration of two or more images of the same scene taken at various times or by various sensors. To reduce the loss of contrast, texture, spectral, and spatial information in the fused image, we suggested an infrared (IR) and visible (VIS) image fusion technique employing a combination of (Latent Low Rank Representation) "LatLRR" and "ResNet." The fused image is more suitable and explicit for human and computer vision perception. Using LatLRR, the source images are initially broken down into saliency and latent low rank parts. Then, using weighted-average method, the Latent Low Rank components are integrated. We present a unique fusion outline based on deep features to overcome the problem of information lost by using RESNET to extract multilayer features for the saliency part as this network has a skip connection which helps to gather information from every layer. After that, the deep features are normalized by nuclear norm, and initial-weight maps are generated. Combining the original weight maps with a Soft-Max operation yields the final-weight maps. At last, by summing the fused Latent Low Rank component and the Saliency content, the fused image will be reconstructed. Both visual and quantitative performance evaluation is done for the proposed method with five existing methods and it is inferred from the results that the proposed method can well reconstruct the fused image without haloes and artifacts. [ABSTRACT FROM AUTHOR]