Muhammad A. Shahin

Color calibration of scanners for scanner-independent grain grading

ABSTRACT Scanner technology is emerging as a cost-effective and robust imaging alternative to camera-based systems in many applications. However, scanner technology is changing so fast that image quality can vary from model to model. It is critical that images scanned with different scanners be brought to a common basis for processing and measurement through a calibration process that eliminates scanner-to-scanner variability. The focus of this research was to investigate scanner-to-scanner variability and develop color correction or mapping functions to allow for machineindependent grain inspection. Various makes and models of scanners were compared for optical and color characteristics. Three different color correction methods wereevaluated: grayscale (GS) transformation, redgreen-blue (RGB) transformation, and histogram matching. All three models of color correction worked within satisfactory tolerance for a multicolor Q60 chart. However, for grain samples of a limited color range, the histogram matchi...

Detection of Hard Vitreous and Starchy Kernels in Amber Durum Wheat Samples Using Hyperspectral Imaging (GRL Number M306)

The Grain Research laboratory (GRl) of the Canadian Grain Com mission (CGC) is committed to developing objective instrumen tal methods for grading grains and quality assessment of grain products. historically, method development at the GRl has been carried out through independent research in the technological areas of conventional image analysis and near infrared (NIR) spec troscopy. NIR spectroscopy has been suc cessfully used to determine moisture and protein contents in grains 1,2 while imaging

Lentil seed size distribution with machine vision

In this study, a machine vision system was developed to determine size distribution of lentil seeds from images of bulk samples. Seed boundaries were separated using morphological processing techniques. The size measurements were partitioned into various size categories in terms of histograms. The system was evaluated on different lentil varieties that differed in both seed size and colour characteristics. The vision results were compared with the caliper and sieve measurements to determine system performance. Results of this study indicate that machine vision can be used to obtain seed size distribution from images of bulk lentils.

Seed Sizing with Image Analysis

Seed size is an important grading factor in pulse grains. Seed processors and canning industry prefer shipments with minimal size variability to those with a wider range of seed distribution. Typically, seed size distribution of a batch of grains is determined by sieving a representative sample from a batch. Image analysis based seed sizing provides a faster, more consistent, accurate and effortless alternative to sieving. The objective of this study was to develop a machine vision system for sizing seeds of various shapes. A flatbed scanner based image analysis application was developed to size circular (peas), elliptical (soybean) and multifaceted (chickpeas) shaped seeds by imaging a bulk poured sample. This application automatically separates the seed boundaries in an image, measures individual seeds, and reports size distribution for user-selectable sieve combination in metric or imperial units. Development of the image analysis system along with its performance in comparison with manual sieving is discussed in this paper.

Use of Imaging Methods for Assessment of Asian Noodle Color

Asia represents a major component of the wheat export market for Australia, Canada, and the United States. Depending on the Asian country, 35–40% of the imported wheat is used for the production of a wide variety of noodles (6). Fresh yellow alkaline and white salted noodles are very popular, and consumer purchasing is initially determined by product appearance (21,22). Noodle appearance can be influenced by flour protein content (23), degree of flour refinement (18), enzyme levels (1), flour particle size and the degree of starch damage (8), the presence of degrading factors (7,13), the type of alkaline reagent used (21,24), enrichment (16), and even the amount of water used in noodle production (9). The most common problem in noodles is speckiness due to the presence of small wheat bran particles in the flour. Wheat bran is a natural component of flour, but its presence and the size of the particles is a function of the degree of flour selection and sieving during the milling process. In many Asian countries, noodles are manufactured daily in small plants and transported to various vendors. The time, elevated temperature, and humidity after production all influence the changing appearance of the noodle product. Bran particles are very rich in a wide variety of phenolic compounds that undergo oxidation by enzymes and yield undesirable colored byproducts (4). Over time (24 hr), large areas of discoloration are produced that cause the consumer to reject the product. Image analysis methods can measure the degradation in noodle appearance. Image analysis of raw noodles can effectively discriminate a large number of factors influencing final product appearance (14). For example, image analysis can effectively discriminate and quantify the degree of bran contamination and its impact on noodle appearance in both major noodle types (12). Initial research utilized costly CCD cameras to detect undesirable features on noodle surfaces. Refinements in imaging technology, in combination with advances in scanner technology, have resulted in the use of inexpensive scanner systems capable of providing noodle manufacturers with a high degree of appearance assessment and discrimination (28). Continuous advancements in the associated software for the analyses of captured images provides manufacturers with the ability to customize their quality control practices to meet the demands of their market niche (15,26). This technology also allows noodle manufacturing firms to stipulate quality specifications to their suppliers. An operator has the ability to quantify noodle appearance in seconds on the basis of number of specks, size of speck, and discrimination from the background matrix. While the use of inexpensive color scanners provides a wide variety of essential noodle quality information critical to the manufacturer for meeting the continually changing needs of consumers, concurrent color assessment from the captured images will greatly benefit the noodle industry. The current procedure for assessing the noodle color under either commercial or laboratory conditions is through the measurement of CIE L*, a*, and b* values on a small piece of noodle, usually <2.5 cm in diameter. A series of readings taken across the entire noodle sheet is averaged to quantify noodle color. The use of commercial colorimeters, while objectifying color measurement, does not offer any means to integrate and measure consumer perception of speckiness. An imaging system could potentially measure noodle color as perceived by humans. The captured images of noodles show the overall noodle color represented by the distribution (histogram) of individual color channels. Histograms have been widely used to represent, analyze, and characterize images for pattern recognition and content-based image retrieval from databases (5,25,30). The fundamental task that remains is to somehow relate image color histograms to the CIE L*, a*, and b* values measured with a colorimeter. Black and Panozzo (2) compared two techniques to predict the color of grains and wheat flour (L*, a*, b*) based on nearinfrared reflectance (NIR) spectra. They found that the standard CIE method for computing color (Standard E308-95) performed better than calibrations derived from the spectra and reference colorimeter data. An earlier study (3) showed that a neural network model for converting a device-dependent color space (red, green, blue [RGB] and hue, light, saturation [HLS]) to a device-independent color space (CIE L*, a*, b*) outperformed statistical as well as optimization approaches in terms of lowest error. Neural networks have been widely used for modeling complex problems where input-output relationships are not readily discernible (17,19,27,29). The objective of this study was to incorporate existing noodle imaging refinements with the development of a fast, reliable, and relatively inexpensive imaging method to concurrently predict the color of Asian noodles.

17 – Quality Evaluation of Corn/Maize

Abstract Corn plays an important role in peoples daily life. Therefore it is very important to detect corn quality. However, traditional quality evaluation methods are tedious, time-consuming, and destructive. Computer vision as a rapid and nondestructive technique has been widely used for corn quality detection and evaluation. This chapter describes the latest applications of corn quality detection based on computer vision, including breakage and cracks detection, quality classification, toxins detection, etc. In addition, emerging imaging techniques such as hyperspectral imaging, which infuses the merits of computer vision and traditional spectroscopy, has the potential to overcome the difficulties of computer vision in terms of corn quality evaluation. Finally, future trends of computer vision on corn quality detection have also been proposed.

Development of multispectral imaging systems for quality evaluation of cereal grains and grain products

Abstract: The commercial value of small grains such as wheat and barley is determined by their overall quality. Grading systems established by the major countries exporting these grains establish maximum tolerances for various contaminants and damage factors to the grain kernels. Grading factors are mainly visual characteristics. Collectively, the grading factors describe grain quality and safety as affected by growing conditions, handling and storage practices. Human visual inspection is the current method of grain quality assessment, which can be subjective and inconsistent. Alternative instrumental approaches to quality assurance are constantly being researched. Hyperspectral imaging has been used effectively for the detection and quantification of grain damage due to various grading factors such as mildew, fusarium, sprouting and green immature seeds. Hyperspectral systems, while effective, are expensive research tools. Multispectral systems using a few wavelengths provide practically viable, less expensive and simpler imaging solutions. A multispectral system using three band-pass filters successfully detected and scored green barley kernels. Development of fast, accurate and low-cost multispectral systems is expected to have a profound effect on the acceptance of instrumental methods of grain grading by the grain industry.

Measurement of spaghetti speck count, size, and color using an automated imaging system.

An objective imaging method was developed to count dark specks in spaghetti. The method simultaneously measured individual speck size and color and the overall color of the spaghetti. Spaghetti samples were prepared from durum wheat samples collected from the Prairie Registration Recommending Committee for Grains (PRRCG) durum wheat cooperative trials during four consecutive crop years from 2002 inclusive to 2005. Differences in speck counts were found between samples within each year. From year to year, the baseline for speck counts varied with the highest numbers in 2005 and the lowest numbers in 2004. For comparison, three technicians also counted the number of specks in each sample. These visual counts were not consistent between technicians or technician to the imaging method, supporting the need for this objective approach. Spaghetti speck counts did not relate to the speck counts of the semolina subsequently used to prepare the product. Speck sizes were consistent across samples and between years, indicating a consistent milling method for all the samples. Differences in speck count numbers could not be attributed to differences in speck color or pasta color. The imaging method gave very consistent speck counts and color measurements over the four years.

Determination of short-term sequential scan variability in a scanner-based machine vision system for grain grading

During routine use of flatbed scanners for grading lentils, unexpected short-term variability in scanner performance was noticed. This variability was detected as scan-to-scan differences, or repeatability. This study aims at developing an objective measure of scanner repeatability to facilitate the selection of scanners by establishing performance criteria for a scanner-based vision system. Four measures of scanner repeatability were compared, using seven scanners for two types of target objects. Scanner selection for grain grading applications is discussed. For grain grading, scanner repeatability is best characterized by color variations as measured by Eq. (4), which performed better than the other three measures of repeatability. The selection of the method of determining scanner repeatability, and the tolerance thereof, are dependent upon the constraints of the imaging targets.