P. O. Box 43, 6700 aa wageningen, The Netherlands
Download 128.8 Kb. Pdf ko'rish
|
demol2001
Detection model for mastitis in cows milked in an automatic milking system R.M. de Mol a,* , W. Ouweltjes b a Institute of Agricultural and Environmental Engineering (IMAG), P.O. Box 43, 6700 AA Wageningen, The Netherlands b Research Institute for Animal Husbandry (PV), Runderweg 6, 8219 PK Lelystad, The Netherlands Received 1 February 2000; accepted 25 November 2000 Abstract Automated detection of diseases (such as mastitis) in dairy cows might be an alternative for detection by observation during milking Ð especially when using an automatic milking system (AMS). An outline of a detection model is given. This detection model includes time-series models for two variables (milk yield and electrical conductivity of milk), with interpolation on previous values. The model is flexible in the number of variables actually used. Parameter values and the residual variances are updated by linear regression after each milking. Alerts for mastitis are given when the residuals fall outside given confidence intervals. A data set with 111 cows for 16 months (on average, 58 lactating cows per day) was used to test the model. Depending on the chosen confidence interval, 42±44 out of 48 cases of clinical mastitis were detected; the remaining cases were not detected because not all data needed were available. These results were better than the results obtained with the model usually used on the farm. The number of false-positive alerts depended on the chosen confidence interval and was higher than the number found with the model usually used. # 2001 Elsevier Science B.V. All rights reserved. Keywords: Cattle-microbiological disease; Mastitis; Detection; Monitoring; Automatic milking systems; Time-series 1. Introduction Detection of diseases is an important element of dairy cow-status monitoring. Early detection of diseases such as mastitis may restrict harmful consequences for the cow and yield losses. The total losses from clinical mastitis were $83 per cow per year in a herd Preventive Veterinary Medicine 49 (2001) 71±82 * Corresponding author. Tel.: 31-317-476459; fax: 31-317-425670. E-mail address: r.m.demol@imag.wag-ur.nl (R.M. de Mol). 0167-5877/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 5 8 7 7 ( 0 1 ) 0 0 1 7 6 - 3 with average risk, and $206 in a herd with twice the average risk (Houben, 1995). Higher demands on the quality of milk also make detection of abnormal milk more important. Hitherto, detection is mostly done by visual observation of the cows in the milking barn during milkings (2±3 times a day) and by inspecting foremilk. Abnormal milk can be withheld from the bulk tank. Detection of mastitis can be automated by using sensor measurements (Frost et al., 1997; Geers, 1994). Mastitis influences the milk composition, resulting in an increased electrical conductivity. Furthermore, the body temperature can be higher and the milk yield can be lower. Automated mastitis detection is based on conductivity measurements combined with measurements of the yield and the temperature of the milk. Currently, commercially available sensors and detection models give alerts for deviations of single variables. The farmer has to combine this information with his own observations and decide what to do. In conventional milking systems, the alerts can be used for management to support the observations in the milking barn. Alerts for visually detectable abnormalities are less useful than early alerts for future problems. Milking can be fully automated by installing an automatic milking system (AMS), which enables an increased milking frequency and milk yield per cow, and a reduced work investment and work load (Artmann, 1997; Rossing et al., 1997). With an AMS, the observations of the milker during milking are no longer available. Alerts for visually detectable abnormalities are essential and early alerts can add extra value. Sensors are involved in the automation process and can also be applied for monitoring. Automated detection is therefore particularly suited for cows milked in an AMS. False-positive alerts are problematic when milk is automatically separated and they also lower the belief of the farmer that they indeed indicate problems. Therefore, a reduction of these false alerts is important. A detection model for cows (milked twice-a-day; TSM2) was developed earlier (de Mol et al., 1999) for a combined processing of the sensor measurements. This model uses conductivity, yield and temperature measurements to generate mastitis alerts. A time- series model is used to calculate expected values for yield, temperature and conductivity. An alert is given when a combination of deviations between expected and actual values is outside a chosen confidence interval. The detection model was tested on two experimental farms (de Mol et al., 1997) and in a field test on four farms (de Mol et al., 1998). With the 99% confidence interval, 90% of all cases were detected on the experimental farms (1.8% of non-deviating milkings were detected falsely) and 67% in the field test (2.1% detected falsely). This model is not suitable for AMS farms. Usually, cows can visit an AMS more-or-less voluntarily. Cows are milked when the interval between subsequent milkings exceeds a chosen threshold (e.g. 6 h); otherwise, they are rejected by the AMS. Thus, the situation on farms with an AMS (relative to farms without AMS) is different in two ways: 1. Milking frequency is variable. Cows in an AMS may be milked more frequently, and more than twice-a-day. Actual frequency depends on the capacity of the AMS and its settings. 2. Milking intervals are more variable. The length of the interval is in between a lower limit and an upper limit. The lower limit depends on the system threshold for 72 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 acceptance by the AMS. Cows are taken to the AMS by the farmer when the upper limit is exceeded (e.g. when the previous milking is more than 24 h ago). These aspects of farms with an AMS have consequences for the variables used in the detection model for mastitis. The detection model for cows milked twice-a-day cannot be used, primarily because a fixed milking frequency and a fixed milking interval are absent. In the twice-a-day model, expected yield is based on the daily yield (estimated by the sum of the last two yields), which, in the AMS, is generally not possible to estimate easily. Furthermore, milk temperature has a diurnal rhythm (higher in the afternoon); current temperature is best compared with the temperature 24 h ago (de Mol et al., 1999). Again, in general, this comparison is not possible in an AMS. Finally, conductivity often shows a diurnal rhythm. In the current project, we developed and tested a mastitis-detection model for cows milked in an AMS. This model was based on a generalisation of a detection model for cows milked twice-a-day. The model described the expected behaviour of cows when they did not have mastitis. Deviations between the actual and expected pattern resulted in alerts for mastitis. The objectives of this research were to develop an adequate detection model for cows milked in an AMS and to test the model with the available data set. The test results were compared with detection results found with other data sets, and with another detection model for the same data set. 2. Material and methods 2.1. Data collection A data set was collected at the Research Institute for Animal Husbandry (PV) in Lelystad, The Netherlands (Table 1). The unit was equipped with a single-stand AMS. The manufacturer of the AMS also delivered a detection model, based on exponential smoothing (Table 2). This model ESx (exponential smoothing model, variable frequency and interval) was used by default on farms with this type of AMS. Mastitis alerts of ESx were based on deviating conductivity values. Details of this model were not available. The cows were kept indoors during the experimental period. On average, there were 58 lactating cows per day. Farm observations of clinical mastitis were recorded. Mastitis detection was based on visual inspection of cows (three times per day), after alerts were Table 1 Measurement period and other characteristics of the data set evaluating mastitis detection in an AMS Measurement Values Start data collection 14 September 1997 End data collection 21 March 1999 Number of cows 111 Number of milkings 83,918 Average number of milkings per cow per day 2.6 Average 305-day milk yield 9550 kg R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 73 given for electrical conductivity, milk yield and milk temperature by model ESx. Cows were also inspected when they did not voluntary access the AMS within expected time limits. Also, the milk filter was inspected; clots or blood in the filter might appear when one or more cows had mastitis. Milk samples of composite milk to determine somatic-cell counts were collected every 3 weeks. Thirty-two reports of bacteriological examinations of milk were available of cows suffering from or suspected of mastitis. This data set was used to test the detection model. 2.2. Model description Model TSMx: a time-series model with variable interval and frequency (Table 2) described the behaviour of the variables and updated the parameters in these models after each milking. For the development of TSMx, a data set was used from the experimental farm of the Institute of Agricultural and Environmental Engineering (IMAG), in Duiven, The Netherlands, collected during an experiment in January till March 1997, from 20 cows (Ketelaar-de Lauwere et al., 1999). 2.2.1. Yield Expected yield, based on the daily yield (in the last 24 h) as in model TSM2, could not be calculated straightforwardly in case of an AMS. A linear function was used to model the cumulative yield in between two successive milkings. Interpolation of this piecewise linear cumulative yield was used to calculate the yield during the last 24 h. An example is given in Fig. 1, where four milkings of a cow are given: the current milking at 18.00 h (yield 10 kg), at 8.00 h (8 kg), at 23.00 h the previous day (8 kg) and at 15.00 h the previous day (7.5 kg). These yields were used to construct a piecewise linear function for the cumulative yield. The interpolated daily yield for the current milking was based on this piecewise linear function. The value of this function at 18.00 h at the previous day is 10.5 kg, so the interpolated daily yield for the current milking is 23.0 kg (33.5ÿ10.5). The interpolated daily yield was modelled by a current level and a trend: the so-called ``local linear-trend model'': Y D t m Y t a Y t t Z Y t (1) where Y D t is the daily yield at time t, calculated by linear interpolation on the cumulative yield; t the time of milking in decimal number of days (e.g. 3.25 is 6.00 h into Table 2 Three detection models mentioned in this paper on the evaluation of mastitis detection in an AMS Model name Based on New or old model a Milking frequency Milking intervals TSM2 Time-series models Old Two times a day Fixed TSMx Time-series models New Variable Variable ESx Exponential smoothing Old Variable Variable a New: developed in research described in this paper. Old: available from earlier research. 74 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 day 3); m y (t) the current level of daily yield at time t; a Y (t) the local trend of daily yield at time t; Z Y (t) the random disturbance at time t. Hidden periodicities in a given time-series could be found by plotting periodograms (Chatfield, 1989). Analysis with periodograms showed no major periodicities in the residuals. For most cows, only a peak at a low frequency (indicating a long-term pattern) was observed. The time since last milking did explain a part of the random disturbances, because long intervals (more than 12 h) resulted in lower yields, consistent with literature (Ouweltjes, 1998). Fitting for mean and trend was enough to remove most underlying patterns; therefore, the remaining residuals could be considered noise. 2.2.2. Conductivity As in model TSM2, the expected conductivity in TSMx was based on the values 0.5 and 1 day earlier. The conductivity 12 and 24 h prior to milking was estimated by linear interpolation on the measurements before and after these times. The conductivity was modelled on the interpolated values by an AR(2) model: E q t ÿ m Cq t a Cq t E q t ÿ 1 2 ÿ m Cq t b Cq t E q t ÿ 1 ÿ m Cq t Z Cq t (2) where E q (t) is the conductivity of quarter q for milking at time t; m Cq (t) the average conductivity of quarter q at time t; a Cq (t) the parameter of the conductivity model for quarter q at time t; E q t ÿ 1 2 the conductivity of quarter q, 12 h ago, calculated by linear interpolation; b Cq (t) the parameter of the conductivity model for quarter q at time t; E q t ÿ 1 the conductivity of quarter q, 24 h ago, calculated by linear interpolation; Z Cq (t) the random disturbance on conductivity of quarter q at time t. Fig. 1. Example of the piecewise linear cumulative yield function () built up from the yield (*) at various milkings of a cow. R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 75 Periodograms of the residuals after fitting did not show structural periodicities, so the conductivity model appeared to be appropriate. 2.2.3. Parameter fitting The parameters in the TSMs for the variables of cows milked in an AMS model (Eqs. (1) and (2)) were not known beforehand. They might be cow-dependent and might change in time, as in model TSM2 (Table 1). The parameters were fitted by an iterative regression procedure. After each milking, the parameters were fitted by linear regression on the milkings up to the latest milking. This type of fitting was only possible when enough measurements were available. A steady-state model was used if the number of measurements was between 5 and 25. In that case, only the average value was fitted. The actual value was compared with the overall average value, if the number of measurements was less than five. The yield model was fitted on the measurements during the preceding 30 days, because the level and trend of yield change during lactation. Once cow-dependent parameter values and the variance of the residuals were known, alerts could be calculated. An alert was given when the combination of the actual residuals fell outside a given confidence interval (as in de Mol et al., 1999): the residual for one of the quarters exceeded a threshold and the sum of the residuals of the conductivity (of that quarter) and yield fell outside the confidence interval. Three confidence intervals were used: 95, 99 and 99.9%. So, in model TSMx, after each milking, the following steps were taken: 1. Calculate the interpolated values of each variable needed in the time-series models (Eqs. (1) and (2)). 2. Calculate the residual of each variable using the parameters based on the measurements up to the latest milking. 3. Generate combined alerts if the values are outside the 95, 99 or 99.9% confidence interval, using the calculated variance based on the residuals up to the latest milking. 4. Calculate updated parameter values by linear regression on each variable, including the latest measurements. 5. Calculate the residual variance including the actual residuals. 2.3. Test procedure The model outcomes (alerts for mastitis) were compared with actual occurrences of clinical mastitis. A case of mastitis was classified as true positive (TP) if one or more alerts were given in a defined period around the recorded date of an observed case, otherwise the case was false negative (FN). This mastitis period comprised the day clinical mastitis was recorded plus the preceding 6 days. The number of TP and FN cases was used to calculate the relative sensitivity, defined as the percentage of TP cases: TP= TP FN 100%. The milkings without an alert outside periods of clinical mastitis cases were not always true negative (TN), because the cow might have subclinical mastitis. Milkings were only classified as TN when the occurrence of subclinical mastitis was very unlikely. For the 76 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 classification of TN milkings, cows were selected without any case of clinical mastitis, with samples of cell counts never exceeding 500,000 cells/ml and no positive results of bacteriological examinations (if any). Milkings with an alert for these cows were false positive (FP). The number of TN and FP milkings was used to calculate the relative specificity, defined as the percentage of TN milkings: TN= TN FP 100%. Sometimes, the models could not draw conclusions from the sensor measurements. These ``indeterminable'' variables could be caused by measurement errors. For yield, indeterminable variables also could be caused by start-up effects (e.g. first milkings in a new lactation). The detection results were influenced by the measurement errors (indicated as indeterminable variables). Mastitis cases with measurement errors were difficult to classify. Sensitivity for clinical mastitis was calculated based on cases without indeterminable conductivity as well as based on all cases. Specificity for mastitis was calculated on milkings without indeterminable conductivity as well as based on all milkings. Cases with indeterminable values of yield were still used in the tests. 3. Results For each case of clinical mastitis, a classification with models TSMx and ESx was determined. All cases without indeterminable conductivity were detected with model TSMx (100% sensitivity); the sensitivity was 66% with model ESx (Table 3). When cases with indeterminable conductivity were included, the sensitivity was 67% for ESx and varied between 92 and 88% for TSMx (sensitivity in Table 3). Twenty-five cows never had mastitis (based on observed cases of clinical mastitis and sampling results of cell counts and bacteriological samples). Mastitis alerts for these cows were classified as FP. The specificity was 99.3% with model ESx; with model TSMx the specificity varied between 87.4 and 97.6%, depending on the chosen confidence interval (Table 4). The specificity was slightly higher when milkings with indeterminable conductivity were considered TN (specificity in Table 4). Table 3 Clinical mastitis detection, found with alerts of the model TSMx with three confidence intervals (% in brackets) and with alerts of the model ESx, based on 48 cases Model Tp a FN b TP c FN d Sensitivity e (%) Sensitivity f (%) TSMx (95) 19 0 25 4 100 92 TSMx (99) 19 0 24 5 100 90 TSMx (99.9) 19 0 23 6 100 88 ESx 23 12 9 4 66 67 a Number of true-positive cases. b Number of false-negative cases. c Number of TP with indeterminable conductivity. d Number of FN with indeterminable conductivity. e Relative sensitivity with cases with indeterminable conductivity excluded sensitivity TP= TP FN 100%. f Relative sensitivity with cases with indeterminable conductivity included sensitivity TP TP = TP FN TP FN 100%. R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 77 4. Discussion 4.1. Detection models Three models were relevant for this research: one new model and two old models for comparisons (Table 2). Model TSM2 was meant for cows milked twice-a-day, and therefore not applicable for cows milked more times a day or in an AMS. Model TSMx was especially developed for cows milked in an AMS with variable frequency and intervals of milking. The global structure of TSMx was similar to TSM2 (de Mol et al., 1999): time-series models for the variables with an update of the parameters after each milking. But, model TSMx was based on interpolated values of previous measurements and the time-series model for yield was different. Model ESx was used by default on the AMS farm. ESx reflected the current practice on the farm. Therefore, results with ESx were included although details of this model were not available. Clinical-mastitis sensitivity results with model ESx were worse than results with model TSMx, while mastitis specificity was better with ESx than with TSMx. On farms with an AMS, detection of all cases of clinical mastitis will have priority for the purpose of controlling udder health. Because milk will be separated automatically, a high specificity is required, however. Because more cases of clinical mastitis were detected, model TSMx will be preferred over ESx to control udder health Ð but both models have a specificity too low for automatic separation. The model ESx also generated alerts for yield and temperature, which might also be false positive. The model TSMx only yielded combined alerts for mastitis. So, the actual difference in the number of FP alerts by ESx and TSMx was smaller than presented in Table 4. More sensors (e.g. body weight) might be available in the near future (de Mol, 2000), which will improve the reliability of the alerts Ð but also will increase the need for a combined processing (as in model TSMx). In follow-up research (de Mol and Woldt, 2001), a fuzzy-logic model was developed for the classification of alerts of model TSMx. Each alert is classified ``true'' or ``false.'' Only alerts that are classified true should be presented to the farmer. The fuzzy Table 4 Mastitis detection, found with alerts of the model TSMx with three confidence intervals (% in brackets), and the model ESx, based on 29,033 milkings of 25 cows without mastitis signs Model Tn a FP b TN c Specificity d (%) Specificity e (%) TSMx (95) 22,729 3278 3026 87.4 88.7 TSMx (99) 24,741 1266 3026 95.1 95.6 TSMx (99.9) 25,487 520 3026 98.0 98.2 ESx 27,861 203 969 99.3 99.3 a Number of true-negative milkings. b Number of false-positive milkings. c Number of TN milkings with indeterminable conductivity. d Relative specificity with milkings with indeterminable conductivity excluded specificity TN= TN FP 100%. e Relative specificity with milkings with indeterminable conductivity included specificity TN TN = TN FP TN 100%. 78 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 classification is a formalisation of the reasoning of the herdsman when he is judging alerts. All cases of clinical mastitis without indeterminable conductivity were classified correctly with the 99% confidence interval. The number of FP alerts was reduced from 1266 (Table 4) to 64 (specificity 99.75%). Indeterminable milkings for model ESx (969 out of 29,033; Table 4) were only caused by measurement errors. The number of indeterminable milkings for model TSMx was 3026. Some data also might have been lost in the process of data collection. The percentage of milkings with measurement errors was 3.3%, which was low compared with results from an earlier study (de Mol et al., 1998). An adequate automated detection on a farm with an AMS will only be possible when the number of measurement errors is almost zero. In case of an alert, the measurement values were still used for updating the parameters after a milking (as explained in Section 2.2), because the deviation might be caused by a systematic change (and not by a case of mastitis). Alert measurements might be excluded in a TP case, but might result in an endless sequence of FP alerts in an FP case (e.g. a change in feeding resulting in a different yield level). Alerted measurements were always used for updating the parameters, because it is not certain whether an alert is TP or FP. The yield model (Eq. (1)) is used to detect only short-term changes in the milk yield. Therefore, it was not necessary to include the lactation curve in this model. The lactation curve must be included to detect long-term deviations in yield. Clinical-mastitis diagnoses were based on farm observations. Observations from the milking parlour were not available in case of an AMS. However, these farm observations might be partly based on conductivity measurements from the model ESx. Our results showed that farm observations were not only based on alerts from the model ESx. Sixteen out of 48 mastitis cases were observed on the farm but not detected by ESx (Table 4). The farm observations were adequate as was checked by a comparison of weighted-average cell counts of the AMS farm with eight other farms of PV (Fig. 2). The levels of cell Fig. 2. The average value of cell count samples (1000 cells/ml) against day number (within the experimental period), on nine farms of PV. R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 79 counts of the AMS farm were comparable with the levels of the other farms where mastitis observations could be based on observations in the milking parlour. The weighted-average cell count values of the AMS farm would be higher if clinical mastitis observations were not adequate. The mastitis frequency on the AMS farm (48 cases in 83,918 milkings) was comparable with the frequency in practice: one case in 2000±2400 milkings (Brand et al., 1996) and with the frequency on other farms of the Research Station (data not shown). 4.2. Perspectives for practical application The mastitis-detection results of the current research were compared to those obtained with model TSM2 found in earlier research on farms where cows were milked twice-a- day (de Mol et al., 1997, 1998). The sensitivity in the current research was higher and the specificity was lower. A comparable sensitivity might be expected while the same conductivity sensors were used, but the different implementation of the sensors in the AMS might be the cause for the improvement. The difference in specificity might be explained by the different structure of the model and the collected data. The objective of automated detection on farms equipped with an AMS is different from farms where cows are milked in a milking barn. In the latter case, a detection model gives additional information, besides the visual observations. No observations during milking are available on a farm with an AMS, so the detection model might be the only way to identify deviating cows. The detection results might be influenced by the absence of temperature measurements. Inclusion of temperature sensors would be expected to improve mastitis detection, as indicated by the study of de Mol et al. (1997), in which alerts based only on conductivity were compared with alerts based on conductivity, yield and temperature. Sensitivity and specificity found in the current research, were better than the estimated sensitivity and specificity found by consultation of experts (Van Asseldonk et al., 1998). The estimated sensitivity and specificity found in that consultation (on a farm with conductivity, yield and temperature sensors) was 71 and 86%, respectively. It appeared that these experts were too pessimistic. Although sensors for electrical conductivity have been available for a number of years, they are used by only a small number of farmers with conventional milking systems. To be useful in conventional milking systems, the mastitis alerts must add something to what the milkers can see with their eyes. All stands have to be equipped with sensors (a considerable investment). For automatic milking, the situation is different: there is no visual inspection during milking, and fewer sensors are needed. As is shown in several studies, not all clinical cases are detected and early warning is not reliable with the sensors and software that are currently on the market. Furthermore, the service costs in conventional milking parlours are currently too high. This means that investment in this equipment is usually considered not worthwhile for conventional milking parlours. Also, in this study we were not always able to generate reliable early warnings Ð but could improve the detection of clinical cases and reduce the number of false alerts. Although this may not be enough to justify investment in a milking parlour, it is very useful for 80 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 automatic milking. The conductivity sensors in an AMS seem to be less fragile, which improves their performance and reduces maintenance costs. Additional information that can also be used for detection purposes (but is not used in the present models) is the number of visits to the AMS and the concentrate feeder, the recorded concentrate leftovers and the occurrences of previous cases of clinical mastitis. It is to be expected that AMS will be equipped with more sensors in the future. Some of the extra information that will be recorded can be useful to detect abnormal milk or abnormalities of the cow. The techniques described in this paper can be used to combine all this data into comprehensive information. The method used might also be useful for oestrous detection with activity sensors. This makes management by exception reality, although it appears to be impossible to reach a sensitivity and specificity both of 100%. This means that there will remain a task for the herdsman in mastitis detection. 5. Conclusions Detection of clinical mastitis on farms with an AMS can be automated and presents an adequate substitute for detection by visual observation in the milking barn during milking. A detection model for cows milked with a variable frequency and intervals of milkings (as described in Section 2.2) can be used. The results are promising: a high sensitivity, 100% (all cases are detected, if enough measurements are available) and a rather high specificity, 98% (in case of a confidence interval of 99.9%). References Artmann, R., 1997. Sensor systems for milking robots. Comput. Electron. Agric. 17, 19±40. Brand, A., Noordhuizen, J.P.T.M., Schukken, Y.H., 1996. Herd Health and Production Management in Dairy Practice. Wageningen Pers, Wageningen, The Netherlands, p. 543. Chatfield, C., 1989. The Analysis of Time Series: An Introduction, 4th Edition. Chapman & Hall, London, p. 241. de Mol, R.M., 2000. Automated detection of oestrus and mastitis in dairy cows. Ph.D. Thesis. Wageningen University, Wageningen, The Netherlands, 177 pp. de Mol, R.M., Woldt, W.E., 2001. Application of fuzzy logic in automated cow-status monitoring. J. Dairy Sci. 84, 400±410. de Mol, R.M., Kroeze, G.H., Achten, J.M.F.H., Maatje, K., Rossing, W., 1997. Results of a multivariate approach to automated oestrus and mastitis detection. Livest. Prod. Sci. 48, 219±227. de Mol, R.M., Ouweltjes, W., Kroeze, G.H., Hendriks, M.M.W.B., 1998. Detection of estrus and mastitis: field performance of a model. In: Zazueta, F.S., Xin, J. (Eds.), Computers in Agriculture 1998. Proceedings of the Seventh International Conference, Orlando, FL, October 26±30, 1998. ASAE, St. Joseph, USA, pp. 865±883. de Mol, R.M., Keen, A., Kroeze, G.H., Achten, J.M.F.H., 1999. Description of a detection model for oestrus and diseases in dairy cattle based on time series analysis combined with a Kalman filter. Comput. Electron. Agric. 22, 171±185. Frost, A.R., Schofield, C.P., Beaulah, S.A., Mottram, T.T., Lines, J.A., Wathes, C.M., 1997. A review of livestock monitoring and the need for integrated systems. Comput. Electron. Agric. 17, 139±159. Geers, R., 1994. Electronic monitoring of farm animals: a review of research and development requirements and expected benefits. Comput. Electron. Agric. 10, 1±9. Houben, E.H.P., 1995. Economic optimization of decisions with respect to dairy cow health management. Thesis. Landbouwuniversiteit Wageningen, The Netherlands, p. 147. R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 81 Ketelaar-de Lauwere, C.C., Ipema, A.H., Metz, J.H.M., Noordhuizen, J.P.T.M., Schouten, W.G.P., 1999. The influence of the accessibility of concentrate on the behaviour of cows milked in an automatic milking system. Neth. J. Agric. Sci. 47, 1±16. Ouweltjes, W., 1998. The relationship between milk yield and milking interval in dairy cows. Livest. Prod. Sci. 56, 193±201. Rossing, W., Hogewerf, P.H., Ipema, A.H., Ketelaar-de Lauwere, C.C., De Koning, C.J.A.M., 1997. Robotic milking in dairy farming. Neth. J. Agric. Sci. 45, 15±31. Van Asseldonk, M.A.P.M., Huirne, R.B.M., Dijkhuizen, A.A., 1998. Quantifying characteristics of information technology applications based on expert knowledge for detection of oestrus and mastitis in dairy cows. Prev. Vet. Med. 66, 273±286. 82 R.M. de Mol, W. Ouweltjes / Preventive Veterinary Medicine 49 (2001) 71±82 Download 128.8 Kb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling