The storage and retrieval of data from the hard disk and transfer speed of data from a microcomputer-based sensor to a personal computer are of critical importance for on-the-go sensing. A neural network-based spectral bands decompression model has been developed to optimize acquisition and retrieval of hyperspectral data. The model decompresses or predicts 201 spectral bands from 25 spectral bands between 407 and 940 nm. The model showed strong correlation between decompressed and actual hyperspectral patterns (coefficient of correlation (r²) equals 0.99 and root mean square error (RMSE) equals 0.0004). The decompressed or predicted hyperspectral reflectance patterns were fed into a neural network-based model that predicts chlorophyll readings. The decompressed hyperspectral reflectance patterns showed good correlation to chlorophyll readings (r² equals 0.89, RMSE equals 1.32 SPAD units). The spectral bands decompression model reduces the number of hyperspectral bands to be downloaded from the spectrometer, stored, or retrieved from the hard disk by 87%.

Scab (Fusarium head blight) is a disease that causes wheat kernels to be shriveled, underweight, and difficult to mill. Scab is also a health concern because of the possible concomitant production of the mycotoxin deoxynivalenol. Current official inspection procedures entail manual human inspection. A study was undertaken to explore the possibility of detecting scab-damaged wheat kernels by machine vision. A custom-made hyperspectral imaging system, possessing a wavelength range of 425 to 860 nm with neighboring bands 3.7 nm apart, a spatial resolution of 0.022 mm²/pixel, and 16-bit per pixel dynamic range, gathered images of non-touching kernels from three wheat varieties. Each variety was represented by 32 normal and 32 scab-damaged kernels. From a search of wavelengths that could be used to separate the two classes (normal vs. scab), a linear discriminant function was constructed from the best R((lambda) ₁)/R((lambda) ₂), based on the assumption of a multivariate normal distribution for each class and the pooling of the covariance error that averaged between 2 and 17%, dependent on wheat variety. With expansion to the testing of more varieties, a two-to-four wavelength machine vision system appears to be a feasible alternative to manual inspection.

Mixed pixels constitute AVHRR image. Evapotranspiration (ET) calculated from AVHRR image certainly includes some mistake information of ground objects which have no meaning as far as ET is concerned. Registration of AVHRR image and high spatial resolution image is carried out. A meteorology-RS combination model is used to calculate the ET of the AVHRR image and of the cropping area of the corresponding high resolution image respectively and the difference of these two ETs is given. The error is analyzed and a corresponding statistical calibration method adaptable to test region is given.

Hyperspectral data provides spectral response information that provides detailed chemical, moisture, and other description of constituent parts of an item. These new sensor data are useful in USDA product inspection. However, such data introduce problems such as the curse of dimensionality, the need to reduce the number of features used to accommodate realistic small training set sizes, and the need to employ discriminatory features and still achieve good generalization (comparable training and test set performance). Several two-step methods are compared to a new and preferable single-step spectral decomposition algorithm. Initial results on hyperspectral data for good/bad almonds and for good/bad (aflatoxin infested) corn kernels are presented. The hyperspectral application addressed differs greatly from prior USDA work (PLS) in which the level of a specific channel constituent in food was estimated. A validation set (separate from the test set) is used in selecting algorithm parameters. Threshold parameters are varied to select the best P_c operating point. Initial results show that nonlinear features yield improved performance.

The performance of a multi-spectral apple inspection station capable of orienting some cultivars, conveying, and presenting apples to a camera at five apples per second is described. Apples are pre-sized and hand placed on the conveying devices to rotate about an axis passing through both the stem and calyx of each apple. An image of each apple is captured at four different wavelengths through a common aperture. Special optics and filters allow simultaneous image capture of apple reflectance for wavelength bands of 540 nm, 650 nm, 750 nm, and 950 nm, each with a bandwidth of approximately 60 nm. As each apple is conveyed laterally and rotated through the camera's field of view, 6 regions of interest representing most of the apple's surface at each wavelength band are captured. The images are processed to segment each defect from the surrounding undamaged tissue and the area of each defect is recorded. Typical defects such as new bruises, bruises on stored apples, scab, sooty blotch, corking, rot, russet, and insect damage are detected. Data is shown quantifying the ability of the inspection station to sort damaged apples into appropriate grades for correct pricing in the processing industry.

Early detection of defects and diseases in fruit helps to correctly classify them and make more adequate decisions about the destination of the product: internal market, export or industry. An early fungi infection detection is especially important because a few infected fruits can disseminate the infection to a whole batch, causing great economic losses and affecting to further exports. Ensure products with excellent quality and absolute absence of fungi infections is particularly important in those batches for long conservation or to be exported. The main objective of this work is to detect the fungi infections before they can be visible. Near Infrared spectroscopy has been employed in this work, because it is a non-destructive technique and can be easily implemented on line due to the high speed and simplicity of the process.

This work shows a new technique as a potential methodology to analysis seed. The technology is the dynamic speckle, phenomena produced by the laser illumination on a biological tissue, in this case the seed tissues. In order to develop this technique, it was necessary to know exactly the effect of the water content in the seed on the results, that was evaluated in this research, by the evaluation of the Inertia Moment of seeds in five levels of moisture (13, 20, 30, 37 and 46%). One of the treatments was the use of pvc film to eliminate the evaporation effect. It was analyzed 450 seeds in five degrees of moisture, with and without film, at three consecutive times, and also in a three times replication. Another experiment was the comparison of dead and alive seeds activity by the Inertia Moment in time. The results showed that the moisture influence the level of activity measured by the Inertia Moment technique, that is an important condition to be controlled on the future experiments, and that it is possible to separate dead to alive seeds by Inertia Moment.

A spectral-based weed sensor was designed and tested in laboratory and in field. The effective sensing area of the sensor was determined by measuring sensor responses when weeds were placed at different grid points in front of the sensor. When multiple weeds shared the effective sensing area with soil, the weed classification rate was above 70%. The classification rate was below 50% for single weeds. Under field conditions, the weed classification rate reach 87%. Variations in sunlight did not affect the performance of the sensor significantly. The effect of shadows on the performance was significant.

The smart sprayer, a local-vision-sensor-based precision chemical application system, was developed and tested. The long-term objectives of this project were to develop new technologies to estimate weed density and size in real-time, to realize site-specific weed control, and to effectively reduce the amount of herbicide applied to major crop fields. This research integrated a real-time machine vision sensing system and individual nozzle controlling device with a commercial map-driven-ready herbicide sprayer to create an intelligent sensing and spraying system. The machine vision system was specially designed to work under outdoor variable lighting conditions. Multiple vision sensors were used to cover the target area. Instead of trying to identify each individual plant in the field, weed infestation conditions in each control zone (management zone) were detected. To increase the delivery accuracy, each individual spray nozzle was controlled separately. The integrated system was tested to evaluate the effectiveness and performance under varying commercial field conditions. Using the on-board differential GPS, geo-referenced chemical input maps (equivalent to weed maps) were also recorded in real-time. The maps generated with this system have been compared with other sensing and referencing systems.

Segmentation of ultrasound images is challenging because of the noisy nature and subtle boundaries of objects in ultrasound images. This paper discusses object segmentation and identification for ultrasound fetal images. The feature space for segmentation consists of information extracted from three sources: gray level, texture, and wavelet-based decomposition. Several texture features, including Laws' texture-energy measures and features based on local gray level run-length, were found useful for segmentation. An unsupervised clustering procedure was used to classify each pixel into its most probable class. Morphological operations were used to remove noisy structures from the original gray level images and to improve the boundaries of the segmented objects. An algorithm was developed to locate objects of interest based on a multiscale implementation of an image transform. Fetal heads were identified and their corresponding measurements are made automatically. The method was tested with a set of clinical images. The resulting images showed clearly the segmented objects. The measurements agreed closely with a sonographer's measurements. The purposed method holds promise for processing and analyzing ultrasound fetal images.

A video image analysis system was developed to support automation of beef quality grading. Forty images of ribeye steaks were acquired. Fat and lean meat were differentiated using a fuzzy c-means clustering algorithm. Muscle longissimus dorsi (l.d.) was segmented from the ribeye using morphological operations. At the end of each iteration of erosion and dilation, a convex hull was fitted to the image and compactness was measured. The number of iterations was selected to yield the most compact l.d. Match between the l.d. muscle traced by an expert grader and that segmented by the program was 95.9%. Marbling and color features were extracted from the l.d. muscle and were used to build regression models to predict marbling and color scores. Quality grade was predicted using another regression model incorporating all features. Grades predicted by the model were statistically equivalent to the grades assigned by expert graders.

The U.S. demand for deboned chicken has risen greatly in the past 5 years, with the expectations that this demand will only continue at an accelerated level. The standard inspection process for bones in meat is for workers to manually feel for bones. It is clear that this time- consuming manual inspection method is insufficient to meet the increasing demand for deboned meat products. Georgia Tech Electrical Engineering faculty and Research Scientists in conjunction with a leading x-ray equipment manufacturer are working together on the development of a system to fuse information from visible images and x-ray images to enhance the accuracy of detection. Currently there are some bones that x-ray systems have difficulty detecting. These are usually relatively thin and are located near the surface of the meat. A primary example is a fanbone (so called because of its shape). We will describe and present results from work geared towards the development of an integrated system that would fuse visible and x-ray information. Significant benefits to the poultry industry are anticipated in terms of reduced processing costs, improved inspection performance and increased throughput through the use of the integrated system to be described. Additionally, generic aspects of the proposed technologies may be applicable to other food processing industries.

This paper summarizes the theory of fuzzy inference systems and its application to plant and weed detection. Two simple examples are presented, both of which discriminate between plants and soil and residue backgrounds in color images based on derived excess green and excess red color indices. The first example shows that a numerical excess red model can be readily replaced with a fuzzy inference system, based on training of red, green, and blue inputs and excess red. The second example shows an arbitrary system of fuzzy inference with excess red and excess green using human preselection, which also gives satisfactory discrimination results.

The shape and the condition of stem are important features in classification of Huanghua pears. As the commonly used thinning and erosion-dilation algorithm in judging the presence of the stem is too slow, a new fast algorithm was put forward. Compared with other part of the pear, the stem is obviously thin and long, with the help of various sized templates, the judgment of whether the stem is present was easily made, meanwhile the stem head and the intersection point of stem bottom and pear were labeled. Furthermore, after the slopes of the tangential line of stem head and tangential line of stem bottom were found, the included angle of these two lines was calculated. It was found that the included angle of the broken stem was obviously different from that of the good stem. After the analysis of 53 pictures of pears, the accuracy to judge whether the stem is present is 100% and whether the stem is good reaches 93%. Also, the algorithm is of robustness and can be made invariant to translation and rotation Meanwhile, the method to describe the shape of irregular fruits was studied. Fourier transformation and inverse Fourier transformation pair were adopted to describe the shape of Huanghua pears, and the algorithm for shape identification, which was based on artificial neural network, was developed. The first sixteen harmonic components of the Fourier descriptor were enough to represent the primary shape of pear, and the identification accuracy could reach 90% by applying the Fourier descriptor in combination with artificial neural network.

We discuss the use of photo-reflectance near the critical angle (PRCA) to monitor small changes of the RI of highly turbid liquids. The theory of the reflectance of a laser beam near the critical angle for an external medium with a complex RI is summarized. The applicability of PRCA to sense highly turbid media is demonstrated experimentally on bovine milk samples. We give experimental results showing the temporal variation of the refractive index (RI) during three different processes in bovine milk: (1) Mechanical stirring, (2) temperature changes, and (3) pH variations around the isoelectric point of the casein micelles (micelle aggregation). RI changes in the order of a few times 1 X 10^-3 are observed during the experiments. The experimental results show that the RI of milk can be used to track physico-chemical changes in time allowing one to measure the time constant of the different process. The design of a compact RI probe for in situ applications is discussed. The miniaturization of such a probe will probably limited by factors other than the loss of sensitivity. A novel angle-of-incidence control which requires only linear displacements of some of the optical components (no rotation) is proposed and shown to be feasible. Such an optical probe may be used in the dairy industry and in general in the food industry or food science research laboratories. It could give additional analytical power to the food scientist, engineer, or technician.

Crop production assessments are extremely valuable because of their economic importance in influencing international trade and national economic policies. Associated with the outstanding problems of food supplies in the world, crop production assessments become more and more important. And because of its good time properties and accuracy, remote sensing is becoming an important method of crop production assessments. And models for the forecasting of crop yields using remote sensing satellite data are studied intensively worldwide. After reviewing the experience gained by other researchers in this field, we selected two models, an NDVI-production regression model and a biomass-estimating model, which might be suitable for the estimation of rice and maize production in Korea-north. Satellite data from 1982 to 1997 were utilized within the models to explain the variation of crop production and obtained good results.

This paper provides a preliminary overview of vegetation cover change analysis using a time series remote sensing data in China. A series of 12 monthly average NDVI data, which is derived from NOAA AVHRR data (January 1990 - December 1990), is used in this study. The research result shows that the vegetation cover change is different along the longitude and latitude during the year. Along the same longitude, vegetation index change is closely corresponded to the seasonal change law. The higher the latitude is, the greater the scope of seasonal change. In summer, there is no clear difference in vegetation index between southern and northern region, but in winter this different is much clear. Along the same latitude the vegetation index change is closely related to precipitation change during the year. The result of greenness classification for each monthly NDVI image also shows that in summer, the high greenness classes occupy most area of China; in opposite, lower greenness classes occupy most of territory of China in winter.

Intention of present work is research the influence of non- polar medium and phosphatidilholine on stability of the macromolecules and hydration of cytohrom-C, tripsine and insulin by use of methods laser Raman and Infrared spectroscopy and isotope H/D exchange. It is shown, that the non-polar environment causes convertible changes of spatial pattern of macromolecules a protein degree of order of macromolecules as a result of which is increased. The presence of water at a system results in a converse effect. At interaction of phosphatidilcholin with the protondonors groups a protein will derivate complexes with a hydrogen bonds. Thereof quantity of aminoacidic oddments which are generatix a polar circuit of a plaited layer is augmented. The outcomes of the analysis of bands of compound tone of water testify to presence in a system of three varieties of water clusters distinguished by frequencies of libration oscillations. It is suspected, that the hydrophobic environment can cause reduction of movability of molecules of water in different clusters.

Raman spectra of hydrocarbon and fluorocarbin zigzag structure molecules are studied. The investigations have been fulfilled for following substances: C_nH_2n+2 (n equals 6, 7, 10), C_nH_2nO₂ (n equals 4, 5, 8, 10, 11, 13, 15, 18), C_nF_2n+1Br(n equals 6, 7, 8, 10, 14), for similar structures: H(CF₂)₁₀H, H(CF₂)₁₀CONH₂, F(CF₂)₅CO₂K and commercial products. The frequency shifts of some modes, corresponding to acoustical and optical vibrations, have been observed. The theory, explaining Raman modes frequency dependence on the length of zigzag molecule, is developed. The presence of characteristic isooctane line in Raman spectra of benzines is established. Molecular structure modification of sun- flower-seed oil as a result of technological preparation process has been observed. The obtained results allow detecting zigzag fluorocarbon and hydrocarbon molecules in media and estimating its length with the help of Raman scattering spectroscopy.