Logo Logo
Switch language to English
Dietrich, Oliver (2012): Etablierung einer neuen Methode zur automatisierten Brunsterkennung beim Rind. Dissertation, LMU München: Tierärztliche Fakultät



Heat detection is a key factor for the economic viability of dairy herds. However, it is becoming more difficult for farmers to detect and use the heat for a successful artificial insemination. There are various reasons for this development. On the one hand, cows are showing progressively fewer behavioral signs of oestrus and, on the other hand, due to economic necessity leading to growing herds, the herd managers do not have enough time for adequate oestrus detection. The result is a considerable loss of profit. For these reasons, the development and use of new and above all workable oestrus detection methods are imperative. The objective of this study was to establish a new heat detection method which allows the farmer to automatically observe his herd on a 24/7-basis. The recommended system is based on the continuous detection of the moving cows in a freestall barn, which involves machine vision. The information received about the positions as well as the locomotion within the freestall barn enables a precise view of the moving patterns. With this system, we analyzed the duration and the intensity of activity as well as the social interactions of cows in oestrus with other members of the herd. As gold standard for the validation of the method, we used an enzyme immuno assay test to measure the amount of progesterone in skimmed milk. As part of the extensive preparation, we selected appropriate video cameras, designed suitable markers and implemented special software. Finally, we used high-resolution GigE-cameras and markers we made ourselves out of fabric-reinforced PVC with printed 2D matrix codes. Our study demonstrated that the new method is able to correctly show the increased activity which is typical of cows in oestrus. The median of the increase in motion activity with regard to a moving average of the 10 days before the oestrus was +430 % for cows in natural oestrus and +338 % for cows in PGF2α-induced oestrus. Overall, the median was +397 %. We found a statistically significant difference (p < 0.001) in increased motion activity between cows in oestrus and cows not in oestrus (median: +153 %). There was no statistically significant difference in increased motion activity between cows in natural oestrus and cows in PGF2α-induced oestrus (p = 0.290). In our study we used ROC analysis to find the optimum between sensitivity and specificity of the method. We identified an ideal threshold value of +225 %. The outcome was a sensitivity of 85 % and a specificity of 83 %. To enhance the efficiency of the machine vision method, we established a completely new parameter. This parameter is used for the detection and evaluation of social interactions between cows in heat and peri-oestrus herd members. We named the new parameter “variance of social interactions”. This measure of scale makes sexual contacts of observed cows (in heat) with other likewise sexually active herd members apparent. The median of the variance of social interactions was 1.598 for cows in natural heat, 1.478 for cows in PGF2α-induced oestrus and 1.52 overall. For cows not in oestrus (in dioestrus or gestating), we found a measure of scale of 1.10. We found a statistically significant difference (p = 0.003) between the cows in oestrus and the cows not in oestrus. There was no statistically significant difference in the variance of social interactions between cows in heat naturally and cows in PGF2α-induced oestrus (p = 0.881). With ROC analysis, we determined a heat detection rate (sensitivity) of 74 %. The specificity was 83 %. By means of a regression analysis, we combined the two parameters of increased motion activity and variance of social interactions. This resulted in a sensitivity of 82 % and a specificity of 87 %. Heat detection in dairy herds with the use of machine vision is not only a functional alternative to the pedometers that have been commonly used so far. With the newly introduced and automatically included parameter “variance of social interactions”, the recommended system offers real additional value for the farmer and can thus help to optimize fertility rates in dairy herds. Moreover, the method provides supplemental data such as length of food intake, lying times, average speed, etc.). This data has the potential to help farmers prevent impending diseases such as lameness or ketosis as well as to indicate problems with animal husbandry at an early stage and therefore provide a contribution to animal-oriented management of dairy herds.