Knut Kvaal - Academia.edu (original) (raw)
Papers by Knut Kvaal
Meat Science, 2006
The EUROP classification system is based on visual assessment of carcass conformation and fatness... more The EUROP classification system is based on visual assessment of carcass conformation and fatness. The first objective was to test the EUROP classification repeatability and accuracy of the national senior assessors of the system in Norway. The second objective was to test the accuracy of the trained and certified abattoir EUROP classifiers in Norway relative to EU Commission's supervising assessors. The third and final objective was to test the accuracy of the EUROP classification system, as assessed by the National senior assessors, for prediction of lean meat, fat and bone percentage and lean meat in relation to bone ratio. The results showed that the repeatability and accuracy of the national senior assessors was good, achieving high correlations both for conformation and fatness. For the abattoir assessors, there were some systematic differences compared to EU Commission's assessors, but these differences were within limits accepted by EU Commission. The relationship between abattoir and national senior assessors was good, with only small systematic differences. This may suggest that there also is a systematic difference between the national senior assessors of the system and EU Commission's assessors. The EUROP system predicted lean meat percentage poorly (R(2)=0.407), with a prediction error for 3.027% lean. For fat and bone percentage, the results showed a fairly good prediction of fat percentage, but poorer for bone percentage, R(2)=0.796 and R(2)=0.450, respectively. The prediction error for fat and bone percentage was 2.300% and 2.125%, respectively. Lean: bone ratio was predicted poorly (R(2)=0.212), with a prediction error of 0.363 lean: bone ratio.
Journal of Chemometrics, 2008
ABSTRACT The effects of digitization and quantification in one- and two-dimensional signals on an... more ABSTRACT The effects of digitization and quantification in one- and two-dimensional signals on angle measure technique (AMT) texture analysis are described in order to find optimal corrections. AMT analysis with varying parameters has been carried out on simulated and real images as well on time-series data. All images and signals are of high resolution in order to study these effects with particularly attention to the low(est), local scales. Quantification effects are studied by multiplying original images or signals by varying scale factor values, K with posterior round-up, while digitization effects were realized by applying various resampling schemes for images and signals using bilinear interpolation. Results lead to general conclusions regarding possible operational regimes for correction of these adverse effects as a function of optimal scale- and sampling-factors during AMT pre-processing. Copyright © 2008 John Wiley & Sons, Ltd.
Journal of Cereal Science, 2000
The quality of baguettes can be evaluated by defined sensory attributes and image analyses. The e... more The quality of baguettes can be evaluated by defined sensory attributes and image analyses. The effect of flour quality, production process (traditional French and industrially modified), mixing and proofing time were studied. Process accounted for 40% of the variation in baguette quality whereas flour quality accounted for 16% of the variation when principal component analysis was applied on the sensory attributes. Baguettes produced using a soft dough and gentle treatment (traditional French process) had a higher sensory score for porosity, elasticity, crispness of crust, crackles on the crust, and porosity and volume as measured by image analysis, than baguettes produced using a stiff dough and rough treatment (modified industrial process). Mixing and proofing time also affected the porosity and area of the cut surface. Porosity, crackles on the crust, glossiness and volume were related to flour quality.
Chemometrics and Intelligent Laboratory Systems, 1998
In order to compute the classical texture measures there is often a need to perform extensive cal... more In order to compute the classical texture measures there is often a need to perform extensive calculations on the images and do a preprocessing in a specialised manner. Some of these texture measures are constructed to estimate specific information. Other texture measures seem to be more global in nature. The techniques presented in this paper define algorithms applied on the raw image without extensive preprocessing. We want to show that mathematical transformations of images on a vectorised form will easily enable the use of multivariate techniques and possibly model several features hidden in the images at the same time. In this paper we will compare five different methods of extracting features from textural images in food by multivariate modelling of the sensory porosity of wheat baguettes. The sample images are recorded from factorial designed baking experiments on wheat baguettes. The multivariate feature extraction methods to be treated are the angle Ž . Ž . Ž . measure technique AMT , the singular value decomposition SVD , the autocorrelation and autocovariance functions ACF Ž . and the so-called size and distance distribution SDD method. The methods will be tested on equal basis and the modelling Ž . of sensory porosity from extracted features is done using principal component regression PCR and partial least square re-Ž . gression PLS . The difference between the behaviour of the methods will be discussed. The results show that all the methods are suited to extract sensory porosity but the AMT method prove to be the best in this case. q 1998 Elsevier Science B.V. All rights reserved.
Analytica Chimica Acta, 2000
The reproducibility of chemical sensors is an issue that has to be handled in order to apply sens... more The reproducibility of chemical sensors is an issue that has to be handled in order to apply sensor-array based instruments for quality control purposes on a routine basis in the laboratory. So far, this problem can only be handled by applying mathematical algorithms to describe the temporal variation of the sensor signal. Sensor drift in a commercial solid state based sensor array device has been investigated to develop a drift compensation algorithm in order to handle the sensor drift. The calibration method uses drift compensation algorithms based on curve fitting of the temporal variation of the sensor signal of calibration samples. This procedure eliminates sensor drift within a single measurement sequence and over several sequences (days, months). It is also demonstrated that the algorithm preserves real features in the data structure of real samples that have been measured. However, this requires calibration samples that are highly correlated in sensor response with the real samples to be analysed in order to make a proper drift correction. : S 0 0 0 3 -2 6 7 0 ( 9 9 ) 0 0 7 8 4 -9
Analytica Chimica Acta, 2004
Univariate multiplicative drift correction and multivariate component correction were applied for... more Univariate multiplicative drift correction and multivariate component correction were applied for recalibration of long-term measurement data acquired with a solid-state gas-sensor array system. The efficiency of the methods was evaluated by classifying recalibrated measurement data using k-nearest neighbor classification and partial least-squares discriminant analysis. For the measurement data in this experiment both multiplicative drift correction and component correction appeared to be useful for recalibration of measurement data from the new gas-sensor array with regard to measurement data acquired with the old replaced gas-sensor array.
Journal of Chemometrics, 2008
Meat Science, 2006
The EUROP classification system is based on visual assessment of carcass conformation and fatness... more The EUROP classification system is based on visual assessment of carcass conformation and fatness. The first objective was to test the EUROP classification repeatability and accuracy of the national senior assessors of the system in Norway. The second objective was to test the accuracy of the trained and certified abattoir EUROP classifiers in Norway relative to EU Commission's supervising assessors. The third and final objective was to test the accuracy of the EUROP classification system, as assessed by the National senior assessors, for prediction of lean meat, fat and bone percentage and lean meat in relation to bone ratio. The results showed that the repeatability and accuracy of the national senior assessors was good, achieving high correlations both for conformation and fatness. For the abattoir assessors, there were some systematic differences compared to EU Commission's assessors, but these differences were within limits accepted by EU Commission. The relationship between abattoir and national senior assessors was good, with only small systematic differences. This may suggest that there also is a systematic difference between the national senior assessors of the system and EU Commission's assessors. The EUROP system predicted lean meat percentage poorly (R(2)=0.407), with a prediction error for 3.027% lean. For fat and bone percentage, the results showed a fairly good prediction of fat percentage, but poorer for bone percentage, R(2)=0.796 and R(2)=0.450, respectively. The prediction error for fat and bone percentage was 2.300% and 2.125%, respectively. Lean: bone ratio was predicted poorly (R(2)=0.212), with a prediction error of 0.363 lean: bone ratio.
Journal of Chemometrics, 2008
ABSTRACT The effects of digitization and quantification in one- and two-dimensional signals on an... more ABSTRACT The effects of digitization and quantification in one- and two-dimensional signals on angle measure technique (AMT) texture analysis are described in order to find optimal corrections. AMT analysis with varying parameters has been carried out on simulated and real images as well on time-series data. All images and signals are of high resolution in order to study these effects with particularly attention to the low(est), local scales. Quantification effects are studied by multiplying original images or signals by varying scale factor values, K with posterior round-up, while digitization effects were realized by applying various resampling schemes for images and signals using bilinear interpolation. Results lead to general conclusions regarding possible operational regimes for correction of these adverse effects as a function of optimal scale- and sampling-factors during AMT pre-processing. Copyright © 2008 John Wiley & Sons, Ltd.
Journal of Cereal Science, 2000
The quality of baguettes can be evaluated by defined sensory attributes and image analyses. The e... more The quality of baguettes can be evaluated by defined sensory attributes and image analyses. The effect of flour quality, production process (traditional French and industrially modified), mixing and proofing time were studied. Process accounted for 40% of the variation in baguette quality whereas flour quality accounted for 16% of the variation when principal component analysis was applied on the sensory attributes. Baguettes produced using a soft dough and gentle treatment (traditional French process) had a higher sensory score for porosity, elasticity, crispness of crust, crackles on the crust, and porosity and volume as measured by image analysis, than baguettes produced using a stiff dough and rough treatment (modified industrial process). Mixing and proofing time also affected the porosity and area of the cut surface. Porosity, crackles on the crust, glossiness and volume were related to flour quality.
Chemometrics and Intelligent Laboratory Systems, 1998
In order to compute the classical texture measures there is often a need to perform extensive cal... more In order to compute the classical texture measures there is often a need to perform extensive calculations on the images and do a preprocessing in a specialised manner. Some of these texture measures are constructed to estimate specific information. Other texture measures seem to be more global in nature. The techniques presented in this paper define algorithms applied on the raw image without extensive preprocessing. We want to show that mathematical transformations of images on a vectorised form will easily enable the use of multivariate techniques and possibly model several features hidden in the images at the same time. In this paper we will compare five different methods of extracting features from textural images in food by multivariate modelling of the sensory porosity of wheat baguettes. The sample images are recorded from factorial designed baking experiments on wheat baguettes. The multivariate feature extraction methods to be treated are the angle Ž . Ž . Ž . measure technique AMT , the singular value decomposition SVD , the autocorrelation and autocovariance functions ACF Ž . and the so-called size and distance distribution SDD method. The methods will be tested on equal basis and the modelling Ž . of sensory porosity from extracted features is done using principal component regression PCR and partial least square re-Ž . gression PLS . The difference between the behaviour of the methods will be discussed. The results show that all the methods are suited to extract sensory porosity but the AMT method prove to be the best in this case. q 1998 Elsevier Science B.V. All rights reserved.
Analytica Chimica Acta, 2000
The reproducibility of chemical sensors is an issue that has to be handled in order to apply sens... more The reproducibility of chemical sensors is an issue that has to be handled in order to apply sensor-array based instruments for quality control purposes on a routine basis in the laboratory. So far, this problem can only be handled by applying mathematical algorithms to describe the temporal variation of the sensor signal. Sensor drift in a commercial solid state based sensor array device has been investigated to develop a drift compensation algorithm in order to handle the sensor drift. The calibration method uses drift compensation algorithms based on curve fitting of the temporal variation of the sensor signal of calibration samples. This procedure eliminates sensor drift within a single measurement sequence and over several sequences (days, months). It is also demonstrated that the algorithm preserves real features in the data structure of real samples that have been measured. However, this requires calibration samples that are highly correlated in sensor response with the real samples to be analysed in order to make a proper drift correction. : S 0 0 0 3 -2 6 7 0 ( 9 9 ) 0 0 7 8 4 -9
Analytica Chimica Acta, 2004
Univariate multiplicative drift correction and multivariate component correction were applied for... more Univariate multiplicative drift correction and multivariate component correction were applied for recalibration of long-term measurement data acquired with a solid-state gas-sensor array system. The efficiency of the methods was evaluated by classifying recalibrated measurement data using k-nearest neighbor classification and partial least-squares discriminant analysis. For the measurement data in this experiment both multiplicative drift correction and component correction appeared to be useful for recalibration of measurement data from the new gas-sensor array with regard to measurement data acquired with the old replaced gas-sensor array.
Journal of Chemometrics, 2008