This PDF file contains the front matter associated with SPIE Proceedings Volume 11061, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

In the emerging field of computational imaging, rapid prototyping of new camera concepts becomes increasingly difficult since the signal processing is intertwined with the physical design of a camera. As novel computational cameras capture information other than the traditional two-dimensional information, ground truth data, which can be used to thoroughly benchmark a new system design, is also hard to acquire. We propose to bridge this gap by using simulation. In this article, we present a raytracing framework tailored for the design and evaluation of computational imaging systems. We show that, depending on the application, the image formation on a sensor and phenomena like image noise have to be simulated accurately in order to achieve meaningful results while other aspects, such as photorealistic scene modeling, can be omitted. Therefore, we focus on accurately simulating the mandatory components of computational cameras, namely apertures, lenses, spectral filters and sensors. Besides the simulation of the imaging process, the framework is capable of generating various ground truth data, which can be used to evaluate and optimize the performance of a particular imaging system. Due to its modularity, it is easy to further extend the framework to the needs of other fields of application. We make the source code of our simulation framework publicly available and encourage other researchers to use it to design and evaluate their own camera designs.¹

Phase-measuring deflectometry is a technique for non-contact inspection of reflective surfaces. A camera setup captures the reflection of a sine-modulated fringe pattern shifted across a screen; the location-dependent measured phase effectively encodes the screen coordinates. As the used fringe patterns are much narrower than the screen dimension, the resulting phase maps are wrapped. The number-theoretical solution uses the Chinese remainder theorem to calculate an unwrapped phase map from repeated measurements with coprime fringe widths. The technique is highly susceptible to phase noise, i.e. small deviations of the measured phase values generally lead to unwrapped phase values with large errors. We propose a modification and show how non-coprime period widths make phase unwrapping robust against phase noise. Measurements with two non-coprime fringe period widths introduce the opportunity to discriminate between “legal” measured phase value pairs, that potentially originate from noise-free measurements, and “illegal” phase value pairs, that necessarily result from noise-affected measurements. Arranged as a matrix, the legal measurements lie on distinct diagonals. This insight not only allows to determine the legality of a measurement, but also to provide a correction by looking for the closest legal matrix entry. We present an experimental comparison of the resulting phase maps with reference phase maps. The presented results include descriptive statistics on the average rate of illegal phase measurements as well as on the deviation from the reference. The measured mean absolute deviation decreases from 1.99 pixels before correction to 0.21 pixels after correction, with a remaining maximum absolute deviation of 0.91 pixels.

This work discusses coded aperture functioning in the visible spectral range. A MURA coded aperture was implemented using laser ablation by evaporation of titanium film from a transparent substrate. The optimal laser parameters to make a high-contrast aperture are shown. The paper also present an algorithm based on periodic correlation which has been developed for image reconstruction. Experiments were carried out with the aperture of rank 37 with a minimal element of 130 μm. The images obtained during the experiment demonstrate the effectiveness of the proposed aperture implementation method and the developed image processing algorithm.

3D scanning is a key process in the fields of robotics and computer vision, and can be used for shape comparison between real-world features and Computer-Aided Design (CAD) models. In order to be beneficial, algorithms must be able to accurately register the 3D point cloud to the model surface. In this paper, we propose a registration method using Correlation Coefficients of the point cloud dimensions along with pose calculated by Procrustes Analysis to provide an ideal global registration for the Iterative Closest Point (ICP) algorithm. The method is analysed in the context of objects that are largely smooth and featureless, with only partial scans captured of the objects’ surfaces, and improves the accuracy of registration in such scenarios. This work has resulted in registrations with more accuracy in cases where a rotational alignment is known but a specific position cannot be identified, than if either ICP or Procrustes are used individually, or when Procrustes is used to provide an initial transformation to ICP. It is shown that by applying the proposed method to partially scanned objects, the Root Mean Square Error (RMSE) is significantly reduced. The method is compared with the SAC-IA alignment algorithm, implemented in the Point Cloud Library (PCL), and the results show 0.4mm RMSE for the proposed method and 24.5mm RMSE for the SAC-IA with ICP. The findings in this work could be used in industrial applications including in-situ robotic repair and inspection of free-form manufactured parts.

In order to enhance the efficiency of quality inspection of Direct Copper Bonded (DCB) structures an optical inspection using a 3D measuring system is conceivable. However, it is a challenging task to use 3D optical measurement techniques for diffuse reflective copper surfaces. This work deals with the optical detection of defects of the copper surface, using multi- and hyperspectral acquisition devices. Over a broad spectral range from the visual spectrum to the short-wave infrared (400 nm - 1700 nm) it is analysed which wavelengths provide good contrast ratios for the detection of flaws. For the inspection of the sample back side, the push-broom imager, operating in the VIS and NIR range (400 nm - 1000 nm), provides the best contrast ratio. An outstanding contrast is reached around 400 nm. Deposited particles on the front side of the DCB substrates are best detected by the filter wheel camera, which is sensitive in the visual and near infrared range. Outstanding contrast is reached with wavelengths around 640 nm. After evaluating the standard deviations of the gray values, it can be shown that defects differ clearly from flawless substrate areas under investigation with light of wavelengths 577 nm, 640 nm and 950 nm. Furthermore, the comparison between certain pixel spectra confirms that significant differences appear at the same three wavelengths. Regarding an automated inspection of defects, it is advisable to shift the pattern projection for the 3D correspondence analysis to the spectral ranges mentioned.

Manual visual inspection is the standard method to evaluate cleanliness, scratches, and digs on optical components according to ISO 10110-7 or MIL-PRF-13830B. However, the limited optical resolution of the human eye turns the inspection of critical samples like high power optics or sensor cover glasses into a time-consuming and tiring task with inherently subjective results. Machine vision systems provide an objective alternative with reliable and reproducible results and additional advantages: Documentation in the form of inspection reports can be generated at no extra cost and data can be collected to analyze typical problems in the manufacturing process. Automated inspection systems need high resolution and well-designed illumination for the inspected part to ensure visibility of smallest imperfections with sizes as small as few micrometers. At the same time, short duty cycles of the inspection system are a key user requirement. Manual-visual is performed in dark field conditions. In automated inspection systems, a dark field optical setup is also advantageous, giving the best imperfection contrast. Generating good dark field conditions when inspecting a curved optical, i.e. specular, surface is not trivial. Using a multitude of illumination angles is necessary for good (and orientation-independent) imperfection visibility, but every additional illumination direction increases the risk of specular reflections into the camera, leading to blind spots in the images. We have developed two approaches to deal with this problem: In the first, reflections are avoided by design, in the second, reflections are treated in data processing. Our commercially available ARGOS system is using a line-scan camera in combination with a rotation stage. The light source is illuminating from almost every direction, except for the direction parallel to the line sensor of the camera itself. This gives excellent imperfection visibility while avoiding direct reflections and maintaining dark-field conditions on rotationally symmetric samples with almost arbitrary curvature. The second approach that has been successfully implemented is suitable also for matrix cameras: Multiple illumination configurations are used sequentially, each generating direct reflections in different portions of the image. These image-sets are then combined into a dark-field image without reflections, by applying suitable filter masks to remove specular reflections from the images. The ARGOS-approach with the line-scan camera has the additional advantage, that very high image resolution is possible by rotating the sample while observing a line aligned with the sample radius. In the standard version, image resolution is up to 250 mega pixels; in the high-resolution version up to 1 giga pixel is possible – in scan times of a few seconds. On the other hand, the requirement to rotate the sample is not ideal for an automated inspection of a batch of samples. In cases when the large resolution is not required, e.g. for smaller elements, using a matrix camera is more efficient, as the acquisition of a sufficient number of images with different illumination conditions takes only a fraction of a second, allowing for short overall inspection times.

Saliency analysis is essential to detect common regions of interest (ROI) in remote sensing images. However, many methods imply saliency analysis in single images and cannot detect common ROI accurately. In this paper, we propose the joint saliency analysis based on iterative clustering (JSIC) method to detect common ROIs. Firstly, the size of superpixel patch is adaptively determined by texture feature. Secondly, color feature and intensity feature are utilized to get initial saliency maps and Otsu is utilized to obtain initial ROIs. Finally, iterative clustering is applied to obtain final ROI with less background inference. Quantitative and qualitative experiments results show that the iterative clustering joint saliency analysis method not only has better performance when compared to the other state-of-the-art methods, but also can eliminate image without ROI. Our contributions lie in three aspects as follows: 1) We propose a novel method to calculate the number of superpixel blocks adaptively. 2) A new joint saliency analysis method is proposed based on color feature and intensity feature. 3) We propose a novel saliency modification strategy based on the iterative cluster, which could reduce the background inference and eliminate images without ROIs.

Unseen object detection problem is known as a semantic matching problem. Thus, a semantic matcher takes two images as an input – the request image and the test image. The request image represents an object class needed to be found on the test image. In this paper, we propose a new region proposal based semantic matcher. In our region based semantic matcher we use the same ideas as in R-CNN. Our Body CNN also generates proposals similar to classical Faster R-CNN, and Head-CNN compares proposals with a request descriptor, extracted from the request image. To extract features from the request image we use Request descriptor CNN. All three CNNs – Head, Body and Request descriptor are trained together, end-to-end for seen class object detection by request and then applied to both seen and unseen classes. We have trained and tested our CNN on Pascal VOC Dataset.

In this paper we propose a new algorithm for image filtering using morphological thickness map. Compared to the other smoothing methods, such as anisotropic diffusion, comparative filters, guided and rolling guidance filters, the benefit of our method is that it natively works with the image structure – thickness map, so it does not depend on the various levels of image noise, lightning conditions and effects. We present the method idea, algorithm itself and various experimental results. The results of the filtering using our algorithm can be widely applied in such image processing tasks as image segmentation, motion analysis, invariant feature transformation, data compression.

Stereoscopic imagers are widely used tools for precise three-dimensional (3D) characterization of various objects in industrial and biomedical applications. Narrow-band spectral imaging significantly increases capabilities of these devices, i.e. allows to analyze spatial distribution of spectral properties as well as to achieve higher image contrast, lower optical aberrations and, therefore, to improve geometrical measurements accuracy. Using spectral stereoscopic images for 3D measurements requires a proper choice of a number, positions and width of spectral bands used for calibration and interpolation of the calculated parameters. The experimental determination of these parameters does not guarantee the optimal choice and may be difficult to implement and time-consuming for a large number of spectral bands. In this paper, we demonstrate that the optical design software can be effective for the computer simulation of calibration, comparison of mathematical models and assigning spectral calibration parameters. We show the possibility to optimize the parameters of multi-spectral geometrical calibration to ensure the required measurement accuracy provided by the stereoscopic system on the stage of its optical design before manufacturing via the design of self-developed prism-based imager. Computer simulation allowed us to compare two camera models and various spectral options (conventional white-light as well as arbitrary number, positions and width of spectral channels) applied to calibration procedure. The results of computer simulation are confirmed by multiple experiments. Proposed approach may be used for estimation of 3D measurements errors caused by image noise, tolerances of optical components, temperature variations and other factors.

Quality control of welded joints is an important step before commissioning of various types of metal structures. The main obstacles to the commissioning of such facilities are the areas where the welded joint deviates from acceptable defective standards. The defects of welded joints include non-welded, foreign inclusions, cracks, pores, etc. The article describes an approach to the detection of the main types of defects of welded joints using a combination of convolutional neural networks and support vector machine methods. Convolutional neural networks are used for primary classification. The support vector machine is used to accurately define defect boundaries. As a preprocessing in our work, we use the methods of morphological filtration. A series of experiments confirms the high efficiency of the proposed method in comparison with pure CNN method for detecting defects.

The paper discusses the video processing methods needed to partially automate the processing of results and conduct the "Morris Water Maze" and "Open Field" experiments to study the behavior of laboratory mice depending on various external factors. The software part of the work was done in the C ++ programming language in the QtCreator editor using the computer vision library OpenCV 3.2.

We present a new iteration for a null-screen based conical corneal topographer, designed to be used as an accessory for currently generation smartphones, also featuring pattern recognition techniques, aimed to reduce the effects of systematic errors present whenever a null-screen test is performed on an actual patient. At their current state, they mostly provide information about recommended starting points for the fitting algorithm, with further capabilities still under development.

The defect detection of nonwoven fabrics is one of the most important steps of fabric quality assurance on production lines. For a long time, fabric defects detection has been carried out manually by human vision with an accuracy of about 60–70%, which not only affects the health of the inspectors, but also has high inspection cost. How to automatically detect crease defects of various forms at a high accuracy has been a challenging task in the field of machine vision. At present, Fourier transform and wavelet transform have been adopted to solve this problem. However, both of them can hardly detect stochastic textured in local region from different scales and directions. This paper adopted a 2D Gabor filter-based method to detect the crease defects, which has tunable angular and axial frequency bandwidth, tunable center frequencies, and could achieve optimal joint resolution in spatial and frequency domain. Firstly the fabric crease images are transformed from the spatial domain to the frequency domain. Secondly the frequency domain images are filtered by the Gabor filter with adjustable central frequency, bandwidth and azimuth, and the frequency domain images of the crease pattern are selected in the frequency domain image. Then they are reversed to the spatial domain. Finally the crease area of nonwoven fabric is obtained by the blob analysis. Experiments conducted on various forms of crease defects have shown that by adopting the proposed method, the nonwoven fabric’s crease defects can be detected effectively and accurately.

To the traditional criteria that determine consumer demand, another factor is added: health and healthy lifestyle, which affects the safety and usefulness of the product consumed. There search focuses on the development of formulations and technologies for the production of cooked sausage using whole sesame seeds containing antioxidant sesamin, and the assessment of its quality. The studies were carried out on a control sample and several samples with different options for adding white and black sesame seeds; organoleptic, physico-chemical and microbiological evaluation of the samples, as well as a qualitative response to sesamin content in sausages were evaluated. The best sample of sausage containing whole white sesame seeds in the amount of 5% was determined. Studies have shown that a natural additive in the form of sesame seeds can significantly enrich the organoleptic properties of sausages. The use of vegetable raw materials is a future direction that will ensure the emergence of sausage products with improved and new unique properties on the market. Products enriched with sesamin should be present in the daily diet of man, as this antioxidant is an oncoprotector and hepatoprotector. Because of its ability to retain vitamin E in cells, it has an indirect antioxidant effect. However, precise control over the formulation and technology of cooked sausage production is not an easy task. In the case of whole sesame seeds, which can be white and black, i.e. with very different colour characteristics, the task is greatly facilitated by the use of optical inspection methods.

Improving the quality of grain products is of great importance, since grains are the basis of the modern man diet. In this regard, one of the most important tasks is to control the grain quality parameters at the stages of its assessment and classification, transportation and storage, processing and use in production. This work is devoted to the study of the influence of the mutual arrangement of grains, their orientation and color on the determination of wheat grains vitreousity using computer vision system. A significant decrease in the calculated value of the vitreousity index at a close arrangement of grains relative to each other was revealed. The color of the grain is also an important indicator. The experimental results showed that the transmittance of red grain is much lower than that of white, due to the presence of pigment cells. This fact makes it impossible to determine the vitreousity of a mixture of different types of wheat without a preliminary color analysis, which greatly complicates the quality analysis process. To overcome the identified limitations, it was proposed to use the IR source as the lower illumination source. The experiment showed that the use of IR illumination allows to completely exclude the influence of the mutual arrangement of grains and their color on the determination of the vitreousity index.

The quality and value of precious stones is determined by a set of indicators such as color, purity, dimension, type and quality of cut. This article is devoted to the problem of assessment of the color of precious stones as one of the crucial quality indicators. To date, the color is usually evaluated by an expert who relies on his/her own color perception and a set of master stones (precious stones color standards). The authors propose an alternative way of determining the color using special hardware and software and standard colorimetric techniques. On the example of green ("emerald") shades the possibility of digital classification of precious stones by color is shown.

The quality of meat and meat products is an extensive range of properties that characterize the nutritional and biological value of food and product characteristics, such as: organoleptic, functional, technological, structuralmechanical, sanitary-hygienic, etc., as well as their degree of manifestation. The value of these indicators, in most cases, depends primarily on the composition of raw materials, biochemical changes that occur in the process of processing, external influences. The food product must contain substances and components that are necessary for the human body for normal metabolism. At the moment there are many ways of raw meat processing. Especially one of the popular meat product is a sausage that is a meat product, with or without casing, consisting of minced meat. Monitoring of the processing procedure conduction as well as product quality is an important task. It is established that the analysis of sausages digital images in the RGB system makes it possible to obtain correlation dependencies between the values of parameters in individual channels - R, G, or B and the most important standard indicators of the quality of meat products, such as pH and the ratio of water and dry matter, which allows us to build an automated quality control system for finished meat products.