Search

Software and Cyberinfrastructure for Astronomy VII (2022), Montreal, Jul 17-22, 2022.--Proceedings of SPIE - The International Society for Optical Engineering vol. 12189 Article number 121891J.-- Full list of authors: Sinjan, J.; Calchetti, D.; Hirzberger, J.; Orozco Suárez, D.; Albert, K.; Albelo Jorge, N.; Appourchaux, T.; Alvarez-Herrero, A.; Blanco Rodríguez, J.; Gandorfer, A.; Germerott, D.; Guerrero, L.; Gutierrez Marquez, P.; Kahil, F.; Kolleck, M.; Solanki, S. K.; del Toro Iniesta, J. C.; Volkmer, R.; Woch, J.; Fiethe, B.; Gómez Cama, J. M.; Pérez-Grande, I.; Sanchis Kilders, E.; Balaguer Jiménez, M.; Bellot Rubio, L. R.; Carmona, M.; Deutsch, W.; Fernandez-Rico, G.; Fernández-Medina, A.; García Parejo, P.; Gasent Blesa, J. L.; Gizon, L.; Grauf, B.; Heerlein, K.; Korpi-Lagg, A.; Lange, T.; López Jiménez, A.; Maue, T.; Meller, R.; Michalik, H.; Moreno Vacas, A.; Müller, R.; Nakai, E.; Schmidt, W.; Schou, J.; Schühle, U.; Staub, J.; Strecker, H.; Torralbo, I.; Valori, G., The ESA/NASA Solar Orbiter space mission has been successfully launched in February 2020. Onboard is the Polarimetric and Helioseismic Imager (SO/PHI), which has two telescopes, a High Resolution Telescope (HRT) and the Full Disc Telescope (FDT). The instrument is designed to infer the photospheric magnetic field and line-of-sight velocity through differential imaging of the polarised light emitted by the Sun. It calculates the full Stokes vector at 6 wavelength positions at the Fe I 617.3nm absorption line. Due to telemetry constraints, the instrument nominally processes these Stokes profiles onboard, however when telemetry is available, the raw images are downlinked and reduced on ground. Here the architecture of the on-ground pipeline for HRT is presented, which also offers additional corrections not currently available on board the instrument. The pipeline can reduce raw images to the full Stokes vector with a polarimetric sensitivity of 10-3 · Ic or better. © COPYRIGHT SPIE., This work was carried out in the framework of the International Max Planck Research School (IMPRS) for Solar System Science at the Max Planck Institute for Solar System Research (MPS). Solar Orbiter is a space mission of international collaboration between ESA and NASA, operated by ESA. We are grateful to the ESA SOC and MOC teams for their support. The German contribution to SO/PHI is funded by the BMWi through DLR and by MPG central funds. The Spanish contribution is funded by FEDER/AEI/MCIU (RTI2018-096886-C5), a "Center of Excellence Severo Ochoa" award to IAA-CSIC (SEV-2017-0709), and a Ram ' on y Cajal fellowship awarded to DOS. The French contribution is funded by CNES.

In this paper, starting from the noise and blur characteristics of the strain clamp X-ray image , an image processing method combining bilateral filter denoising and unsharp mask enhancement is proposed. Experimental results show that unsharp mask algorithm have certain limitations in image edge enhancement. After the image is denoised by bilateral filtering,the unsharp mask method is used for image edge enhancement, which can effectively expand the scope of the unsharp mask algorithm. This method greatly improve the image qualityand help identify defects. It provides a strong guarantee and support for the X-ray inspection of the strain clamp, and has important practical value.

Generative adversarial network (GAN) has become a hot research topic in the field of image processing. As an unsupervised training model, GAN has been widely used in the field of computer vision, especially in image style transfer. The purpose of the GAN is to make the generator generate a false image, and the discriminator cannot tell whether the input image is the real image or the generated image. Compared with traditional network models, GAN model has these advantages in image style transfer: GAN is composed of two different networks, and the loss function is automatically learned by playing games with each other. GAN belongs to unsupervised training and does not need to annotate the data set, which saves a lot of work. In this paper, improved GAN models related to image style migration are summarized. Firstly, the principle and method of image style transfer based on convolutional neural network are introduced. Secondly, the status, principle and prospect of GAN are introduced, and the causes of gradient disappearance and mode collapse of GAN are analyzed in detail. On this basis, the principles, advantages and disadvantages of CGAN, DCGAN, CycleGAN and StarGAN V2 network models are introduced. Finally, it summarizes the current problems and future research directions of style transfer based on GAN.

A method for measuring the incident angle of the light beam based on an axicon lens which defined by cone length and cone angle is presented in this letter. The system is mainly composed of a stop, an axicon lens, a telescopic system, an image receiving device, a display device and a digital image processing device. The light beam is refracted to a larger incident angle when passing through the telescopic system. It then goes through the aperture and the axicon lens, whose refraction times in axicon lens correlate with the incident angle, resulting in different spots captured by image sensor. The angle can be figured out with rich image characteristic information, which includes symmetry axis direction, area, the mass position center, the inertia moment, etc. In a practical design example, the image processing of the spot image shows that the rotational inertia of the spot has a one-to-one correspondence with the incident angle of the light beam, which features an obvious monotonicity. The angle measurement range can reach up to 3°, with an accuracy up to 1×10-3 ". Compared with the traditional angle measurement method, the measurement accuracy of this method based on axicon lens is improved by an order of magnitude. The system has the advantages of high accuracy, compact but simple structure, and no cumbersome installation and debugging. It can be used not only as a high-precision collimator, but also as a high-precision goniometer.

Fused silica optical elements are widely used in large high-power laser devices, and the subsurface defects of optical elements directly affect the laser damage threshold and imaging quality. These defects are distributed within a few to tens of microns from the surface and cannot be detected by conventional imaging methods. The characteristics of small size, low density and wide distribution range make it difficult to detect nondestructively, effectively and quickly with conventional methods. In order to solve the above problems, an experimental system was built based on the principles of photoluminescence and dark field scattering. Use the different characteristics of fluorescence image and scattering image to carry out the research of subsurface defect detection. First, preprocess the original image. Second, calculate the offset of adjacent images, perform background homogenization processing on the image, and stitch the sub-aperture images, and then segment the full-aperture image. Then, a more effective subsurface defect extraction algorithm is proposed. Finally, etch the sample with HF and observe it under a microscope. Experimental results show that this method can detect weak defects on the surface and sub-surface at the same time, and can effectively separate individual sub-surface defects. This method has the advantages of lossless, fast and high precision. Moreover, based on a large amount of data, analysis and summary of the causes of defects, distribution characteristics, etc.. This method can provide certain guidance for the evaluation of laser damage threshold and processing technology of optical components in high-power laser devices.

With the development of visual human-computer interaction, non-invasive eye-tracking technology has been applied in medical diagnosis, psychological research, augmented reality and other fields. As a key step of video-based eye tracking technology, pupil detection requires strong robustness, real-time performance and high precision. Although many pupil detection algorithms have been developed, pupil detection is still a challenge when there are so many interference factors such as bright light, reflection, eyelid or eyelash occlusion, or low image resolution. Aim to address the above difficult problems of pupil detection, this paper uses cascaded Haar features and Otsu dynamic threshold segmentation to improve the performance of the existing pupil detection algorithm. The performance of the improved algorithm is verified with the public human eye image dataset from Swirski with high off-axis and severe eyelash occlusion. The results show that the accuracy of the improved algorithm is 86.3% and 89.7% in the error range of 5 pixels and 10 pixels, respectively. The detection algorithm in this paper can provide accurate pupil center coordinates for eye tracking, which lays a foundation for the high-precision calibration and measurement of the following eye tracker.

Aiming at the images of relevant monitored objects in the process of laser cladding, a super resolution algorithm technology was proposed to optimize and enhance the key details of the images, and the enhanced image content was segmented, extracted and counted. First, construct a training of a sub-resolution convolutional neural network (SRCNN) model; the original low resolution is predicted by the weight after training, the image quality evaluation results: peak signal-to-noise ratio (PSNR) is 30.198212, structural similarity (SSIM) is 0.969966; the most based on the maximum entropy dual threshold split algorithm combined with image processing, extracting and statistics on the powder object in the segmentation result image, the number of effective powders and proportion of the original wandering map and the predicted output delay image is [112, 33.6%] and [240, 40.6%]. The research results show that the cladding image output from the original image after the super resolution model has been significantly improved in terms of clarity and quality as well as the optimization and enhancement of the details of the monitored object.

Phase retrieval is of great significance in the fields of medical imaging and computational holography. To solve the problem more efficiently, this paper presents a phase recovery network with two modes of operation based on the convolution neural network, which not only can get persistent model by training data set, but also can build a special loss function to recover the unknown signal in a self-optimized way without data set in the case of Gaussian measurement model. Comparison of the simulation results show that the network is able to obtain better results with fewer measurements than the existing phase recovery algorithms.

Quality inspection is an essential technology in the glass product industry. Machine vision has shown more significant potential than manual inspection at present. However, the visual inspection of the bottle for defects remains a challenging task in a quality-controlled due to the difficulty in detecting screw thread defects. To overcome the problem, we propose a bottle mouth detection method based on area segmentation. First, an area segment method with traditional image processing methods, which is based on the characteristics of screw thread, was proposed. According to the result of the segment, the bottle area is divided into a screw thread area and a non-screw thread area. For the former, a defect detection method that uses edge detection and Gaussian filters is proposed to precisely detect screw thread defects. For the latter, a defect detection method that uses techniques such as the Sobel algorithm and global binarization is proposed to precisely detect other defects. The proposed method is tested for data sets obtained by our designed vision system. The experimental result demonstrates that our framework achieves good performance.

A superpixel consists of a series of small regions composed of pixel points that are located next to each other and have similar features such as color, luminance, and texture. Most of these small areas retain the original information of the image, which facilitates faster follow-up processing. Most of the existing superpixel segmentation methods do not use deep learning network architectures. Some methods use deep learning, but they also have a very simple network architecture. And there are a few superpixel segmentation methods combined with deep learning. In this method, a more complex convolutional neural network with an attention module is applied to superpixel segmentation. The resulting superpixel segmentation network is more complex, and the addition of the attention module allows for more accurate chunking of the images, thus yielding more comprehensive and detailed segmentation results. By experimenting on a public dataset BSDS500, the method has higher accuracy in superpixel segmentation. Also, the segmentation speed of the method is similar to that of the existing simple segmentation networks.

There are many mountainous areas in my country, and the geological disasters caused by the sliding and collapse of slopes threaten road traffic safety at all times, and seriously hinder social and economic development. This research is designed to realize image processing technology and slope sliding monitoring system. It proposes a fusion template corner detection algorithm based on fusion template, improves Harris corner detection algorithm, and solves the problem of traditional algorithm's large amount of calculation and dependence on response threshold setting. In addition, there are shortcomings such as corner clustering, and then experiments are used to verify the effectiveness of the improved algorithm in the extraction of feature corners of slopes and mountains.

This paper proposes a large depth-of-field (DOF) imaging method, which uses an image acquisition system equipped with a liquid lens. This paper also establishes the relationship between the focusing distance and the current value to estimate the object distance. The proposed method firstly uses a liquid lens to capture multi-focus images. Then the different objects in the scene are segmented by image processing algorithm. Then we apply an effective blur estimation algorithm on each object to determine the clearest image from all multi-focus images and obtain the corresponding current value. Finally, the clear images of different objects are combined to obtain a large DOF image, and meanwhile, the distances of different objects can be estimated, respectively. Experimental results show that the proposed method can effectively achieve large DOF imaging and measure approximate object distances.

In the field of ocean and seismic wave detection, the distributed acoustic sensing based on chirped pulse is used to obtain the specific information of different waves and to dig out the available effective content. But with the huge amount of data, there will be some problems of low computational efficiency and long processing time. This paper applies image processing methods to analyze the phase velocities of different waves. The method combines different technologies to get the related analysis more quickly. Specifically, on the basis of the original data, the two-dimensional Fourier transform processing is performed on the two-dimensional matrix as preprocessing, and then the edge detection algorithm and the operation of color inversion is used to process the data of 20-30 kilometers. After that, all the results are converted into scatter plot form. Then the required scatter points need to be kept and fitted, and this step contributes to obtaining the range of phase velocity variation of different waves through the tangent of the fitted curve. The final result is in line with the phase velocity of the actual situation analysis, and the process is more convenient. The methods proposed in this paper are more convenient to calculate, and they reduce the workload. As a result, they are suitable for monitoring the phase velocity of seismic wave and ocean wave detection.

Image denoising is an important topic in the field of image processing. With the application of nonlocal similarity in sparse representation, the work of image denoising began to be performed on similar patch groups. The sparse representations of patches in a group will be learned together. In this paper, we propose a novel image denoising model by combining group sparsity residual with low-rankness. Firstly, motivated by the relationship between low rank and sparsity, a low rank constraint is imposed on the sparse coefficient matrix of each similar patch group to enhance the sparsity. Secondly, since 𝛾-norm can most closely match the true rank of a matrix, it is applied for rank approximation in our model. Finally, in view of the fact that numerous iterations are required in the group sparse representation (GSR) model, we develop an efficient algorithm based on the Majorize-Minimization (MM) optimization. It greatly reduces the computational complexity and the number of iterations. Experimental results show that our model makes great improvements in image denoising and outperforms many state-of-the-art methods.

Information on flower number per grapevine inflorescence is critical for grapevine genetic improvement, early yield estimation and vineyard management. Previous approaches to automize this process by traditional image processing techniques, are unsuccessful in the improvement of a universal system that can be applied to inflorescences with diverse morphology. In this paper, we validated the efficiency of advanced deep learning-based approaches for automatic flower number estimation. The results were analyzed on an inflorescence dataset of 204 images from four different cultivars during various growth stages. The algorithm developed on patch-based instance segmentation using Mask R-CNN, produced counting results highly correlated to manual counts (R2 = 0.96). Practically constant MAPE values among different cultivars (from 5.50% to 8.45%), implying a high robustness in this method. Achieving the fastest counting (0.33 second per image of size 512 × 512) with slightly lower counting accuracy (R2 = 0.91), the method based on object density-map using U-Net turned out to be suitable for real-time flower counting systems.

Underwater image is an important carrier of marine information. However, the problems of low contrast, color degradation, uneven illumination and detail loss cause the degradation of image quality. In this paper, we design a novel underwater image enhancement pipeline based on retinex. Firstly, we use bilateral filter instead of the traditional gaussian filter to estimate illumination image and obtain the reflectance image. Then, we design an attenuation map guided gray world method to overcome color distortion and use gamma correction to improve the illumination image. Finally, the fusion image is post-processed to further improve the contrast and obtain the final enhanced image. Qualitative and quantitative performance analysis prove that the proposed method has better performance than other underwater image enhancement methods.

With the development of computer vision and deep learning, the convolutional neural network has been widely used in image processing such as object detection and semantic segmentation, and has achieved breakthrough achievements. However, when the training samples are insufficient, the conventional neural network usually has unsatisfactory robustness. In order to solve the problem, we improve the generalization performance of the few-shot detectors by focusing on the target center and can identify novel categories. The paper proposes a new attention mechanism based on the auxiliary circle feature map of the object center. By selecting an auxiliary circle feature map with the object center as the center of the circle and the minimum size in height and width as the diameter, adding it to the anchor-free CenterNet network as soft attention to promote network training. Several experiments on PASCAL VOC2007/2012 datasets show that the proposed method achieves the most advanced level in terms of the accuracy and standard deviation of few-shot object detection, which indicates the algorithm’s effectiveness.

Recently, the functional extension of optical coherence tomography (OCT) with OCT angiography (OCTA) allows volumetric imaging of the subsurface microvasculature without requiring exogenous contrast agents like conventional angiography techniques. However, performing OCTA requires intensive computation to extract the changes of OCT signal due to moving red blood cells in the microvascular network. In this study, we have developed a graphic processing unit (GPU)-accelerated framework to realize high speed OCTA imaging and the visualization of the microvascular network after the data acquisition. In addition, we investigate the feasibility of providing real-time microvascular imaging leveraging dynamic scattering OCT and GPU.

Line detection in images has always been one of the important research fields of artificial intelligence computer vision for image processing.Hough transform (HT) is one of the most extensive detection algorithms in image processing and machine vision applications, and has the advantages of anti-noise. However, it involves a huge amount of computation and too much memory.In view of its shortcomings, it is recommended to improve its performance. In this paper, a new detection method is cited, and the statistical probability Hough line transformation is added.The first step is to use the Canny operator to obtain the edge information of the gray image and use the improved method to detect the straight line segment in the image. This algorithm enhances the accuracy of edge detection and optimizes the extraction of the straight line. At the same time, it has the simplicity and wide applicability of Hough transform.

The feeding volume parameter of combine harvester greatly affects harvesting efficiency. Therefore, this paper proposes a feeding volume prediction method. This method uses the camera installed at a certain height above the front of combine harvester to capture the images of crops during operation. Then, the ARM processor unit does image processing and BP neural network inference. Here, the image processing includes color space conversion, histogram equalization, filtering operation and other operations. And its output is the pixel value of rice panicle layer. As for the neural network, the pixel value, rice stubble height, moisture content and grass to grain ratio are used as the input parameters, and the output of network is feeding volume. The result of neural network shows the decisive factor of predictive model up to 0.95, which is able to predict the feeding volume properly. At last, according to the result of experiments, this method predicts feeding volume well, and the relative error of the predicted average feeding amount is 10%.

Contour detection of an object is a fundamental computer vision problem in image processing domain. The goal is to find a concrete boundary for pixel ownership between an OOI (object-of-interest) and its corresponding background. However, contour extraction from low SN SEM images is a very challenging problem as different sources of noise shadow the estimation of underlying structural geometries. As device scaling continues to 3nm node and below, the extraction of accurate CD contour geometries from SEM images especially ADI (after developed inspection) is of utmost importance for a qualitative lithographic process as well as to verify device characterization in aggressive pitches. In this paper, we have applied a U-Net architecture based unsupervised machine learning approach for de-noising CD-SEM images. Unlike other discriminative deep-learning based de-noising approaches, the proposed method does not require any ground-truth as clean/noiseless images or synthetic noiseless images for training. Simultaneously, we have also attempted to demonstrate how de-noising is helping to improve the contour detection accuracy. We have analyzed and validated our result by using a programmable tool (SEMSuiteTM) for contour extraction. We have de-noised SEM images with categorically different geometrical patterns such as L/S (line-space), T2T (tip-to-tip), pillars with different scan types etc. and extracted the contours in both noisy and de-noised images. The comparative analysis demonstrates that de-noised images have higher confidence contour metric than their noisy twins while keeping the same parameter settings for both data input. When the ML algorithm is applied, the contour extraction results would have higher confidence numbers comparing with the ones only applied the conventional Gaussian or Median blur de-noise method. The final goal of this work is to establish a robust de-noising method to reduce the dependency of SEM image acquisition settings and provide more accurate metrology data for OPC calibration.

Background: Machine learning is predicted to have an increasingly important role in semiconductor metrology. Prediction intervals that describe the reliability of the predictive performance of machine learning models are important to guide decision making and to improve trust in deep learning and other forms of machine learning and artificial intelligence. Image processing is an important application of artificial intelligence. Low-dose images from the scanning electron microscope (SEM) are often used for roughness measurements such as line edge roughness (LER) because of relatively small acquisition times and resist shrinkage, but such images are corrupted by noise, blur, edge effects, and other instrument errors. LER affects semiconductor device performance and the yield of the manufacturing process. Aim: We consider prediction intervals for the deep convolutional neural network EDGENet, which was trained on a large dataset of simulated SEM images and directly estimates the edge positions from a SEM rough line image containing an unknown level of Poisson noise. Approach: Conformal prediction is a relatively recent, increasingly popular, rigorously proven, and simple methodology to address this need for both classification and regression problems, and it does not use distributional assumptions such as Gaussianity or the Bayesian framework; one new variant combines it with another technique to generate prediction intervals known as quantile regression. Results: We illustrate the strengths and limitations of different conformal prediction procedures for the EDGENet approach to LER estimation. Combining these approaches into ensemble schemes and incorporating domain knowledge produces more informative prediction intervals. Conclusions: Deep learning models can help in the estimation of LER, but their acceptance has been hindered by a lack of trust in these techniques. Prediction intervals that provide coverage guarantees are an approach to alleviate this problem and may catalyze the transition within semiconductor manufacturing to a wider acceptance and implementation of machine learning.

Fluorescent molecules play an important role in many fields due to their characteristic of high sensitivity and easy operation. The traditional method for detecting fluorescent molecules is Laser Scanning Confocal Microscopy, but it has light pollution and can only detect a single class of fluorescent molecules at once. Recently, Microscopic Hyperspectral Imaging(MHSI) technology has been used to the detection of fluorescent molecules due to its high spectral resolution and non-destructive detection. However, the low spatial resolution of MHSI makes it difficult to conduct high-precision molecular research. Therefore, it is important to develop an image processing method to improve it. In this work, a twostep data processing method was proposed to enhance automatic classification effect of fluorescent molecules. We used a microscopic hyperspectral system to image the mixed five kinds of fluorescent molecular samples in transmission mode. The first step is based on the difference in unit slope of spectra curve (wavelength range was chosen from 410 nm to 550 nm) between fluorescence molecule and the background, and an image segmentation method based on minimum light transmission point is proposed. The second step is to calculate the relative absorbance of each voxel according to the nearest background voxel found on the basis of image segmentation, and to take the absorbance as the final classification feature. Compared with the traditional transmittance feature on six kinds of machine learning classification models, the average classification accuracy of the new features can be improved by 2.2%, and the time consumption per classification can be reduced by 1/3 approximately. In conclusion, the proposed two-step data processing method is suitable for the classification of multi-kinds of fluorescent molecules, which has the advantages of high efficiency and accuracy, and is expected to be widely used in biology, medicine, materials and other fields.

Biological tissue is a multiple random scattering medium. The study of the propagation of acousto-optic signals in biological tissues is an important and complex issue in acousto-optical tomography(AOT). In this paper, the finite element simulation software COMSOL Multiphysics is used to simulate the propagation of acousto-optic signal modulated by ultrasound in double-layer tissue. The effects of different types of ultrasounds on acousto-optic signals in tissues are studied. The influence of the optical properties of the target tissue and non-target tissue on the acousto-optic signal in the double-layered tissue is also discussed. The simulation results show that the waveform of the acousto-optic signal is very similar to that of the ultrasonic wave. The acousto-optic signal presents a periodic variation as the ultrasonic frequency changes. The peak-to-peak value and average value of the acousto-optic signal are affected by the optical properties of double-layer tissue. However, the modulation depth of the acousto-optic signal only depends on the optical characteristics of the tissue in the ultrasound focal zone (target tissue), and has nothing to do with the optical characteristics of the tissue outside the ultrasound zone (non-target tissue). The modulation depth has a good antiinterference performance, which is conducive to image processing and reconstruction in AOT.

In the dual-camera phase retrieval method, the phase is solved by positive- and negative- defocusing images obtained through a single exposure after dual cameras are installed on an inverted microscope. However, due to the installation error of the cameras, translation and rotation of images exist between the images, resulting inaccurate phase retrieval results. In this paper, we proposed a dual-camera phase retrieval method based on fast adaption image restoration and transport of intensity equation. Firstly, let the positive-defocusing image be the reference image. Then using the fast adaption image restoration algorithm to find the texture information in order to find best matching block quickly. According to the number of high frequency information of the block, the size of block can be defined in order to increase the precision and speed of the restoration. After that, priority can be change as the sum of two parts, which can avoid the situation of 0 priority. Then, burring the boundary point of restored image in order to reduce the block effect. Finally, the transport of intensity equation can be used in phase retrieval results. Comparing with the normal algorithm, this method can restore the image much better.

ABO blood type automatic recognition can realize the automatic detection of blood type, effectively improve the speed and accuracy of blood type analysis, and has a wide range of clinical applications. In this paper, an automatic blood type recognition algorithm based on image processing is designed. Firstly, the image of blood type card is preprocessed, including image enhancement and median filtering. Then, the micro column tubules are segmented by template matching. Then, the threshold analysis is carried out, and the gray image is transformed into a binary image. Finally, the distribution of red blood cell aggregates in micro column tubules is identified, and determine the ABO blood type of the sample. The experimental results show that the algorithm can effectively segment the micro column tubules on the blood type card, realize the effective recognition of red blood cell distribution, and complete the effective interpretation of blood group.

With the rapid change of society and the continuous development of science and technology, the application of instruments is more and more widely in all kinds of places from daily life to chemical power and aerospace. The pointer instruments in traditional industries are still widely used. But without changing the original equipment structure, the intelligent transformation is still an urgent research direction. The object of this paper is a kind of pointer pressure instrument used in industry. The purpose of this paper is to research the reading system of the pointer instrument based on embedded hardware platform and simple software algorithm. This paper designs an automatic reading system of the value indicated by the dial of the pointer instrument. The system uses embedded hardware devices and adopts the design idea of Linux + Qt + OpenCV. An open source, cross platform and relatively simple image processing algorithm is designed in it. Canny Edge detection, Hough-Circles detection and the Hough-Lines detection are combined in it.

In order to ensure that the nondestructive testing climbing robot can track the weld more accurately and in real time, a method of weld edge detection and centerline extraction is designed. The histogram-based method is used to achieve fast median filtering of the image, and the background and target weld are segmented by adaptive thresholding. To obtain the contour information of the weld, the binarized image is subjected to secondary filtering and area filling combined with the opening operation of morphology and the erosion operation. After Canny edge detection and Hough transform linear detection, the straight line equations on both sides of the weld are obtained, and the centerline position of the weld is calculated. Experimental result show that this method greatly reduces the amount of calculation, shortens the image processing time, and can accurately identify the center of the weld.

Nowadays, Quick Response (QR) code technology is used in all aspects of our lives due to the widespread use of smartphones. However, when the Light Emitting Diode (LED) display is suspended from the ceiling, the actual camera is not facing the LED and must have a certain pitch angle when shooting, so the captured image must have a three-dimensional spatial rotation effect, which leads to low decoding rate. Therefore, to address this phenomenon, this paper proposes an image processing method based on the combination of Hough transform and affine transform, which solves the problems of image rotation in the recognition of QR code images at the receiving end.

In geostationary Earth observation satellite system with high-resolution optical system under conceptual study phase developed by Japan Aerospace Exploration Agency (JAXA), one of its main products for users is a video with one fps captured by the telescope and the optical sensor; however, there have been few examples of satellite video with high-rate fps, especially in geostationary satellite. A satellite video has instability of the video sequences caused by an undesired disturbance of the satellite. In this study, we propose the video stabilization method corresponding to various satellite imagery characteristics, using feature point matching and bundle adjustment with eliminating outliers by using RANdom SAmple Consensus (RANSAC). Moreover, we adopt a polynomial approximation to smooth global motions of all video frames. We demonstrate the video stabilization using the experimental video captured by the helicopter. Moreover, the performance test result shows that the proposed method can stabilize the satellite video less than one pixel accuracy. In addition, we discuss various issues specific to a satellite imagery, for example, moving cloud.

Today, almost all imaging systems include both an optical part and an image processing part in order to improve the final image quality. It therefore seems natural to optimize them simultaneously to obtain the best possible result. However, even if this “co-design” approach is more and more recognized at a conceptual level, it is still rarely used in practice for designing complex lenses with many ajustable parameters and constraints. This is due to the fact that the contribution of image processing is currently difficult to take into account in optical design software. Until now, the field of co-design has thus mainly focused on simpler imaging systems, consisting for example of single co-optimized optical elements such as phase masks. More recently, Robinson and Stork have been working on the possibility of integrating the image processing criterion known as mean square error (MSE) in the optical software Zemax OpticStudio. It is also possible to consider surrogate criteria instead of this MSE, built from more classical optical criteria (modulation transfer function, point spread function, etc.) [Burcklen et al. (2018)]. In this study, we investigate the possibility of implementing the MSE criterion in the CodeV optical design software in a way that is easily usable by an optical designer. We compare systems co-designed with this approach to systems jointly optimized with surrogate criteria or conventionally optimized. We focus on the performance differences between these different approaches and on the opportunities offered by CodeV for co-design.

A new approach to obtaining three-dimensional cutting edges on a tapered cutting surface is proposed based on an image obtained on a measuring machine in the control section. The method is developed based on forming a spline and evaluating the shape using a unique algorithm, which is in turn based on an analytical mathematical model. The developed method allows estimating the main geometric parameters of the cutting edge and analyzing the shape of the helical groove of the drill. In the experiments, the drill was installed on a rotary table capable of adjusting the position around its axis. The validity of the new approach was corroborated by physical tests and analysis of the set of images obtained using an optical measuring system.

We report on the development and field trials of an active polarimetric imager in the SWIR domain. Polarization states are controlled for both emission and analysis. Based on past experience, we focus on Orthogonal State Contrast (OSC) imaging for which two images with orthogonal polarizations are needed. An important feature of the imager is the use of two InGaAs imaging detectors mounted orthogonally on a polarization beam splitter. This allows the synchronous imaging with the two orthogonal polarizations and the real time acquisition of OSC images at video frame rate without temporal artefacts. The demonstrator has been operated during field trials with static and moving scenes. These trials were mainly aimed at the detection of man-made objects (weapons, vehicles …) in complex scenes at up to a few hundreds of meters. Along with the presentation of some example of results, we discuss different representation modes of the polarimetric information.

The use of data obtained by a group of sensors requires the formation of parallel channels. The use of each separate channel leads to the need to allocate additional computing resources and the appointment of time intervals (timing) for a single-processor analysis system. The formation of the decision rule and the subsequent decision-making based on such data requires forming a combined inter-block criterion. This criterion should consider both the possible intersection of data and their discrepancy associated with the use of different parameters when processing the same data. The formation of combined data reduces the computational costs at the decision-making stage, which will improve the efficiency of post-processing and visual control systems. In using combined stationary systems, it is possible to create template fields that allow you to form transformation matrices for a specific space. If it is impossible to use fixed cameras or combined systems in a single body, forming stitching images is complicated. Combining data into a single information field also allows you to increase the operator's work efficiency, allowing you to analyze the entire process as a whole and not its scattered parts. The paper proposes a technique for forming a stitching thermal imaging image based on combined data analysis. For the formation of anchor points for stitching images, primary analysis methods with combined processing parameters are used. Images obtained from the outputs of thermal imaging cameras are pre-processed by the filtering method. The method used is based on the application of a multicriteria function. It automatically divides the image into regions (boundaries, highly detailed, and locally stationary areas) to reduce noise while preserving the transition boundaries. To increase the processing speed, a simplification algorithm is applied while maintaining the shapes and geometry of objects. The operation includes absorbing small objects and averaging for the ranges of histograms of color gradients. Analysis of local features and the formation of anchor points is based on the use of correlation analysis. As a method of non-linear change in color balance, modified alpha-rooting methods are used. As test data, a series of images of one object obtained from different fixation points in: visible (camera with a resolution of 1920 x 1080 pixels, color depth 8 bits) and far-infrared (thermal images with a resolution of 320 x 240 pixels, grayscale). Images have at least 40% overlap area of one object. Applications for both industrial production and the analysis of objects in the open areas are considered.

The identification and localization of laser threats in a free-space channel is an increasingly important capability as active laser sensors become more widely used in civilian and military applications. Early identification of a laser source requires o-axis detection of laser light scattered out of the optical channel by molecules and aerosols in the free-space channel. In this work we report on the development of a sensor system for the o-axis detection of scattered laser light from a non-cooperative laser source. An image processing approach to the detection and localization of a laser in a scene is presented. The performance of the sensor system is quantified with experiments performed inside the laboratory and in realistic daylight background conditions.

Synthetic Aperture Radar (SAR) imaging systems operate by emitting radar signals from a moving object, such as a satellite, towards the target of interest. Reflected radar echoes are received and later used by image formation algorithms to form a SAR image. There is great interest in using SAR images in computer vision tasks such as classification or automatic target recognition. Today, however, SAR applications consist of multiple operations: image formation followed by image processing. In this work, we train a deep neural network that performs both the image formation and image processing tasks, integrating the SAR processing pipeline. Results show that our integrated pipeline can output accurately classified SAR imagery with image quality comparable to those formed using a traditional algorithm, showing that fully neural network based SAR processing pipeline is feasible.

Maritime surveillance systems’ long-range capabilities are dependent on the quality of their images. Poor weather conditions such as haze, fog, or smog can severely hamper the ability to observe a thermal imaging scenario accurately, which affects the ability to detect, identify and track any object of interest within it. An image processing technique, known as dehazing is required in these types of systems. In this paper, state of the art image dehazing algorithms are used for long range thermal images (~5km, ~9km, ~15km) and their image restoration quality performances are compared based on qualitative metrics such as structural similarity (SSIM), peak signal to noise ratio (PSNR), and Feature similarity (FSIM). The results from this benchmark study can provide a suitable dehazing technique for maritime surveillance systems.

A central research topic in medical image processing is the development of imaging biomarkers, i.e. image-based numeric measures of the degree (or probability) of disease. Typically, they rely on segmentation of an anatomical or pathological structure in a radiological image, followed by quantitative measurement. With much of traditional image processing methods being supplanted by machine learning techniques, the identification of new imaging biomarkers is also often made with such techniques, in particular deep learning. Successful examples include quantitative assessment of Alzheimer’s disease and Parkinson’s disease based on brain MRI data, as well as image-based brain age estimation.

Energy weighting is an efficient technique for improving image contrast. If finally evaluating through human visual information, it is fascinating how the contrast of the image changes due to the difference in the weighting of X-ray energy. Therefore, we tried to evaluate how X-ray energy and contrast affect each other by comparing with gamma correction, which is a contrast adjustment performed in general image processing. Since the interaction in X-ray imaging is complicated, We replaced the imaging system with a simple one, and the change in the image due to the difference between energy-integrated and photon-counting detectors is represented as a tone curve. Simulation is a suitable technique for calculating the change in pixel value due to the difference in the weight of X-ray energy. Still, it requires creating a model for each imaging target. Therefore, to estimate the change more easily, we examined how to simplify the imaging system.

The root hump defect is easy to appear in the process of high power laser welding. Through the observation of welding experiment, it is found that there is obvious correlation between root hump defect and the character of keyhole and molten pool. Therefore, this paper proposes a method to monitor the root hump defect by identifying the keyhole and molten pool features in the welding process. In this method, image sensing technology and machine vision method are used to analyze and extract the keyhole and weld pool information in real time. The BP neural network algorithm is used to classify the welding states. It is found that adding the feature of weld pool length as input will greatly improve the recognition accuracy of the model.

Lidar is an optical technology for detection and range measurements, which has been attracting much attention in developments of next generation driving and navigation. Herein, we report on development of a type of non-scanning, non-solid-state, laser-based infrared lidar system with the potential applications in advanced driver assistance systems and autonomous vehicles. Our emphasize in the design approach has been on compactness of the final system in order for it to be deployable both as standalone or complimentary to existing lidar sensors, enabling fusion sensing in automotive or even in drone applications. The non-scanning lidar system, currently patent pending, is comprised of a laser light source, a plurality of optical elements to create a predefined reference optical pattern, means for filtering returned optical signals, imaging optics, an optical detector, and a processing unit. The principle of this system is largely based on image processing with a known and calibrated reference. Using Python and OpenCV, the near infrared images acquired from the entire field of view is analyzed in real-time to determine the position and velocity of the objects. The work presented here describes the principle of our new lidar system, the optical system design, as well as the experimental results demonstrating its performance. The benefits and limitations of the imaging lidar technology developed by us are compared to those of the current scanning and flash lidars.

The Philippines used to be one of the prime exporters of coffee from different parts of the world. However due to lack of technology and the absence of standard the production and exportation of coffee diminish through the years. Until now, the coffee farmers are relying on manual operations of classifying and profiling coffee beans intended to level and match the global standard. Hence, the researchers created a system that will automatically classify and profile coffee beans without human intervention based on the different features of coffee beans using integrated image processing algorithms. The focus of this research is to create a device that can evaluate the size, quality, and roast level of a batch of the coffee beans through the use of image processing techniques and Back Propagation Neural Network. To determine these features, BPNN would serve as the method to develop the brain of the device. The integrated processing algorithms used in this research include K-mean shift, Blob, and Canny Edge to extract the features of the coffee beans and Red Green Blue Analysis, Hu's Moment, and Blob Analysis to make use of these features and feed it into the BPNN. Based on the standard set by the Philippine Coffee Board Inc., the prototype in this research was able to classify and profile different coffee beans with up to 100% accuracy.

Infrared imagery, like almost any other two-dimensional (2D) imagery, have been traditionally sampled and acquired using a traditional rectangular grid. Therefore, nonuniformity correction (NUC) algorithms for infrared imaging systems which mitigate the most dominant, bias/offset portion of the nonuniformity were developed on the rectangular grid. However, it is well-known that hexagonal sampling grid captures more information in sampled data/imagery when compared to traditional rectangular sampling, and a hexagonal addressing scheme (HAS) for hexagonally-sampled imagery to convert imagery between the two different coordinate systems was developed. In this work, we build on prior work by Sakoglu et al. who developed bilinear interpolation equations between two image frames under the 2-D global motion of the scene or the camera, and apply this 2D algebraic NUC algorithm to hexagonally-sampled imagery directly in the HAS domain by utilizing simulated hexagonal sampling of real IR images.

Automatic edge detection in images is an area of great interest to industry and the scientific community. A problem usually experienced is that edge detectors are sensitive to the magnitude of changes in brightness. However, this disadvantage disappears when employing the technique known as phase congruency, which allows edge detection in an image regardless of its illumination level. This technique is based on phase alignment of frequency components. This principle states that the edges of an image occur when the phases of the Fourier components coincide. By using phase, the direct dependence on brightness intensity in edge detection is avoided. A difficulty of phase congruency implemented with monogenic filters is that it requires the computation of the complex Fourier transform. However, its computational cost is high. There are some approaches that seek to reduce its cost, such as the FHT, but they only allow to obtain the Fourier transform of real images. Due to this limitation, methods based on phase congruency were not available in several image processing tools, such as ImageJ, a program widely used by biologists and microscopists for the analysis of biological images, since these programs make use mainly of the Fast Hartley Transform transformation. Therefore, in this work the implementation of phase congruency for ImageJ with monogenic filters using the Fourier Radix-2 FFT transform is described. The results obtained with the proposed implementation were compared with those found with the Kovesi code in GNU Octave, showing that both implementations obtain equivalent results and even better with the proposed method when at least one side of the images is not a power of two, in which case tile-mirror is used to complete the image.

An algorithm for the recognition and tracking of several objects in color image sequences is presented. First, each three-channel color input image are encoded into a single-channel complex-valued image. Next, a set of prespecified targets are recognized and located in the scene by a composite matched filtering with complex constraints. Afterwards, the set of targets are tracked by adapting the matched filtering to each input image and by processing small image fragments extracted at the predicted coordinates of the targets in the scene. Results obtained with the proposed algorithm in a test image sequence are presented and analyzed in terms of efficiency of target recognition and accuracy of target tracking.