Detection and Classification of Potato Diseases Potato Using a New Convolution Neural Network Architecture

Using machine vision and image processing as a non-destructive and rapid method can play an important role in examining defects of agricultural products, especially potatoes. In this paper, we propose a convolution neural network (CNN) to classify the diseased potato into five classes based on their surface image. We trained and tested the developed CNN using a database of 5000 potato images. We compared the results of potato defect classification based on CNN with the traditional neural network and Support Vector Machine (SVM). The results show that the accuracy of the deep learning method is higher than the Traditional Method. We get 100% and 99% accuracy in some of the classes, respectively.

Aerial tracer particle distribution system for surface image velocimetry

A novel aerial tracer particle distribution system has been developed. This system is mounted on an Unmanned Aerial Vehicle (UAV) and flown upstream from where surface velocimetry measurements are conducted. This enables surface velocimetry techniques to be applied in rivers and channels lacking sufficient natural tracer particles or surface features. Lack of tracers is a common problem during low flows, in lowland rivers, or in artificial channels. This is particularly problematic for analysis conducted using Particle Image Velocimetry (PIV) techniques where dense tracer particles are required. Techniques for colouring tracer particles with biodegradable dye have also been developed, along with methods for extracting them from Red Green Blue (RGB) imagery in the Hue Saturation Value (HSV) colour space. The use of coloured tracer particles enables flow measurements in situations where sunglint, surface waves, moving shadows, or dappled lighting on riverbeds can interfere with and corrupt results using surface velocimetry techniques. These developments further expand the situations where surface velocimetry can be applied, as well as improving the accuracy of the results.

Smoothing Complete Feature Pyramid Networks for Roll Mark Detection of Steel Strips

Steel strip acts as a fundamental material for the steel industry. Surface defects threaten the steel quality and cause substantial economic and reputation losses. Roll marks, always occurring periodically in a large area, are put on the top of the list of the most serious defects by steel mills. Essentially, the online roll mark detection is a tiny target inspection task in high-resolution images captured under harsh environment. In this paper, a novel method—namely, Smoothing Complete Feature Pyramid Networks (SCFPN)—is proposed for the above focused task. In particular, the concept of complete intersection over union (CIoU) is applied in feature pyramid networks to obtain faster fitting speed and higher prediction accuracy by suppressing the vanishing gradient in training process. Furthermore, label smoothing is employed to promote the generalization ability of model. In view of lack of public surface image database of steel strips, a raw defect database of hot-rolled steel strip surface, CSU_STEEL, is opened for the first time. Experiments on two public databases (DeepPCB and NEU) and one fresh texture database (CSU_STEEL) indicate that our SCFPN yields more competitive results than several prestigious networks—including Faster R-CNN, SSD, YOLOv3, YOLOv4, FPN, DIN, DDN, and CFPN.

SOLVING THE PROBLEM OF DETERMINING THE LOCATION OF AN UNMANNED AERIAL VEHICLE IN THE COORDINATE COUNTING MODE USING A TECHNICAL VISION SYSTEM

Представлено обоснование использования оптико-электронной системы в качестве навигационно-измерительного комплекса. Проведен краткий анализ существующих систем навигации, применимых для беспилотного летательного аппарата, и предложен алгоритм обеспечения системы видеонаблюдения в режиме счисления координат с помощью системы технического зрения. Задачу счисления координат БЛА с использованием видеопоследовательностей изображений земной поверхности можно решить с высокой точностью с помощью бинокулярной СТЗ. Однако в случае выхода из строя одной из камер определение координат местоположения будет продолжаться с достаточной точностью для решения поставленной задачи. А недостаток измерительных средств обеспечивается за счет использования 6 особых точек земной поверхности. Поэтому предложен алгоритм определения местоположения с помощью монокулярной системы технического зрения. Для решения задачи определения местоположения выделяются и определяются координаты особых точек на изображении поверхности. Для нахождения особых точек была выполнена обработка оцифрованного изображения методом FAST-9. Так как изображение получается цветным, то процедура нахождения особых точек является надежным путем применения метода FAST-9 для двух или даже трех цветовых компонент. Данная процедура позволяет достигнуть высокой точности определения счисляемых координат БЛА. Для решения задач счисления координат предпочтительно использование методов простых итераций, Брауна или Ньютона We present the rationale for the use of an optoelectronic system as a navigation-measuring complex. We carried out a brief analysis of existing navigation systems applicable to an unmanned aerial vehicle and propose an algorithm for providing a video surveillance system in the reckoning mode using a vision system. The problem of reckoning UAV coordinates using video sequences of images of the earth's surface can be solved with high accuracy using a binocular TVS. However, in case of failure of one of the cameras, the determination of the coordinates of the location will continue with sufficient accuracy to solve the task. And the lack of measuring instruments is ensured through the use of 6 special points of the earth's surface. Therefore, we propose an algorithm for determining the location using a monocular vision system. To solve the problem of determining the location, we selected and determined the coordinates of the singular points on the surface image. To find the special points, we processed the digitized image using the FAST-9 method. Since the image is obtained in color, the procedure for finding special points is reliable by applying the FAST-9 method for two or even three color components. This procedure allows you to achieve high accuracy in determining the reckoning coordinates of the UAV. To solve problems of reckoning coordinates, it is preferable to use the methods of simple iterations, Brown or Newton

Comparison of Dental Surface Image Registration and Fiducial Marker Registration: An In Vivo Accuracy Study of Static Computer-Assisted Implant Surgery

This study compared the accuracy of static computer-assisted implant surgery (sCAIS) planned through dental surface image registration and fiducial marker registration. Stone models of 30 patients were converted into digital dental casts by using a desktop scanner. Cone-beam computed tomography (CBCT) was performed and superimposed to the digital dental casts with two methods: matching the dental surface images or matching the fiducial markers on a stereolithographic radiographic template. Following the implant planning, stereolithographic surgical guides were fabricated, and 56 fully guided implants were inserted by the same doctor. Deviations between planned and inserted implants were measured and compared using postoperative CBCT images. After adjustment for other potential influencing factors, compared with the fiducial marker registration group, significantly larger mean lateral deviations were noted in the dental surface registration group at both the implant platform and apex (p = 0.0188 and 0.0371, respectively). However, the mean lateral deviations for the dental surface registration (0.83 ± 0.51 mm at implant platform and 1.24 ± 0.68 mm at implant apex) were comparable to the literature. In conclusion, our findings indicate that although sCAIS planned using dental surface image registration was not statistically as accurate as that using fiducial marker registration, its accuracy was satisfactory for clinical use.

Variant of digital image processing application for recognition of textures on the optoelectronic image

Problem statement. Search of plants on the observable image is important function of the UAV onboard optoelectronic systems. One of methods of the observable surface image analysis is the textural analysis which allows to select homogeneous areas on the observable image and to define contours of the plants which are covered with various structures and not having boundaries accurately defined on the image.The purpose. To develop algorithm of the textural analysis of a site terrestrial or a water surface, fulfilled by the digital image processing. In paper the image textural analysis algorithm fulfilled on the basis of an evaluation of local spectral and statistical performances of the image is presented.Results. The algorithm according to which for each element of the analyzed image in a square environ from 1024 elements the local spectra oriented under different angles and local variances are calculated is developed. Calculated by digital image processing amplitudes of a spectrum and variances values characterize texture parameters in a environ of each element of the image. Points of considerable modifications of local properties define position of boundaries between various structures.The practical importance. Results of the considered algorithm trials as a part of simulation model have shown a possibility of its application in the UAV onboard image observed analysis system.