1. |
- Ardö, Hakan, et al.
(author)
-
Convolutional neural network-based cow interaction watchdog
- 2018
-
In: IET Computer Vision. - : Institution of Engineering and Technology (IET). - 1751-9632 .- 1751-9640. ; 12:2, s. 171-177
-
Journal article (peer-reviewed)abstract
- In the field of applied animal behaviour, video recordings of a scene of interest are often made and then evaluated by experts. This evaluation is based on different criteria (number of animals present, an occurrence of certain interactions, the proximity between animals and so forth) and aims to filter out video sequences that contain irrelevant information. However, such task requires a tremendous amount of time and resources, making manual approach ineffective. To reduce the amount of time the experts spend on watching the uninteresting video, this study introduces an automated watchdog system that can discard some of the recorded video material based on user-defined criteria. A pilot study on cows was made where a convolutional neural network detector was used to detect and count the number of cows in the scene as well as include distances and interactions between cows as filtering criteria. This approach removed 38% (50% for additional filter parameters) of the recordings while only losing 1% (4%) of the potentially interesting video frames.
|
|
2. |
|
|
3. |
|
|
4. |
- Guzhva, Oleksiy
(author)
-
Computer vision algorithms as a modern tool for behavioural analysis in dairy cattle
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- Looking at modern dairy production, loose housing, i.e. free stalls became one of the most common practices, which, while widely implemented along with different management routines, do not always include the adjustments necessary for assuring animal welfare. The analysis of interactions occurring between cows in dairy barns and their effect on health and performance is of great importance for sustainable, animal-friendly production. The general aim of this thesis was to investigate the possibilities and limitations of computer vision approach for studying dairy cattle behaviour and interactions between animals, as well as take a first step towards the fully automated system for continuous surveillance in modern dairy barns. In the first study, a seven-point shape-model for describing a cow from the mathematical perspective was proposed and investigated. A pilot study showed that the proposed Behavioural Detector based on the developed shape-model provided a solid basis for behavioural studies in a real-life dairy barn environment. The second study investigated a classification case from the industry: how animal distribution and claw positioning in specific areas could affect the maximal load on floor elements. The results of the study provided more substantial background data for determining the dimensioning of the strength of the slats. The third study aimed to take the first step towards an automated system (so-called WatchDog) for behavioural analysis and automatic filtering of the recorded video material. The results showed that the proposed solution is capable of detecting potentially interesting scenes in video-material with the precision of 92,8%. In the fourth and final study, a state-of-the-art tracking/identification algorithm for multiple objects with near-real-time implementation in crowded scenes with varying illumination was developed and evaluated. The algorithms forming the multi-modular WatchDog system and developed during this project are the crucial stepping stone towards a fully-automated solution for continuous surveillance of health and welfare-related parameters in dairy cattle. The proposed system could also serve as evaluation/benchmark tool for modern dairy barn assessment. Keywords: dairy cattle, image analysis, Precision Livestock Farming, computer vision, deep learning, convolutional neural networks, social interactions, tracking, cow traffic
|
|
5. |
- Guzhva, Oleksiy, et al.
(author)
-
Cow claws on concrete slats in the waiting area in a dairy barn estimated by use of Image analysis
- 2016
-
Reports (other academic/artistic)abstract
- Slatted concrete floors are commonly used in dairy barns for aisles, feeding and waiting areas. Maximum slot opening in Sweden is 35 mm with a maximum of 28% opening area for adult cattle in order to provide the adequate claw support. The construction of the slats has to consider this together with the length of the slats and the load from the weight of the animals on the slats. Presently, the dimension of the load strength of slats is based on assumptions and experience. An alternative approach is to estimate the true load of the animals on the slats by observation of animal distribution on slatted floors. The purpose of this study was to investigate possibilities of using machine learning algorithms and image analysis for assessing actual distribution of animals in the areas of interest and maximal weight load per slat element per unit of time. Images for the study were acquired from three surveillance cameras placed in the ceiling above the common waiting area (size 6x18 m) with entrances to four automatic milking systems (AMS). Then images were used to train a convolutional neural net classifier to detect and locate the cows in the images. Then, a probability distribution of where the claws might be located was constructed. By using this distribution in a Monte Carlo simulation, a probability distribution of the number of claws on each slat could be estimated, and from that, a worst-case estimate of the actual weight load was constructed. Results indicate that the 95% percentile number of claws on 160 mm wide slat area (slat width including the opening) was estimated to 3.03 and on a 560 mm slat area width was 5.63. Cows mounting was found in 7 of 9215 (0.2 %) examined pictures. The method proposed in this report was promising and for this purpose and could be used for practical assessment of animal distribution and loading and thus be a part of the dimensioning of construction.
|
|
6. |
- Guzhva, Oleksiy, et al.
(author)
-
Cows on concrete slats of the waiting area in a dairy barn estimated by use of image analysis
- 2016
-
Conference paper (other academic/artistic)abstract
- Slatted concrete floors are commonly used in dairy barns for aisles, feeding and waiting areas. Maximum slot opening in Sweden is 35 mm with a maximum of 28% opening area for adult cattle in order to provide the adequate claw support. The construction of the slats has to consider this together with the length of the slats and the load from the weight of the animals on the slats. Presently, the calculation of strength of slats assumes construction of multiple slat units instead of single beams. There is presently no use of empirical data on the distribution of animals’ claws on the surface to estimate the load of the animals on the slat beams. The purpose of this study was to investigate possibilities of using machine learning algorithms and image analysis for assessing actual distribution of animals in the areas of interest and maximal weight load per slat element per unit of time. Images for the study were acquired from three surveillance cameras placed in the ceiling above the common waiting area with entrances to four automatic milking systems (AMS). Then images were used to train a convolutional neural net classifier to detect and locate the cows in the images. Then, a probability distribution of where the hooves might be located was constructed. By using this distribution in a Monte Carlo simulation, a probability distribution of the number of hooves on each slat was estimated, and from that, a worst-case estimate of the actual weight load was constructed.
|
|
7. |
- Guzhva, Oleksiy, et al.
(author)
-
Development of a seven-point shape model for analysis of social interactions in dairy cattle
- 2016
-
Conference paper (peer-reviewed)abstract
- The efficiency of an automatic milking system (AMS) relies on a passage rate of cows through the waiting area to the AMS units. In order to assure high number of visits per cow per AMS unit it is important to understand the hierarchy and social interactions between cows that might occur during the time in the waiting area prior to milking. The competition between cows of a different rank to enter the AMS unit could result in a number of aggressive interactions, which could compromise the individual performance and endanger their health and welfare. The aim of the study was to investigate social interactions between cows in the waiting area with four AMS units and collecting data by three top-down view video cameras. A seven-point shape model for each cow was developed to manually tag cows in video sequences. Furthermore, the interactions between pair of cows were annotated into one of five states: body pushing, head butting, head pressing, body sniffing and no interaction. A classification system, investigating features from pair of cow shape models, was developed. The identification of social interactions using this system and crosscheck with ethogram containing descriptions of all the interactions was tested on a set of video sequences. The system showed a potential for further development and algorithm flexibility dependent on a number of features involved in the analysis. For the single frame analysis system showed higher accuracy with ‘line border features’ than without: 83.1 vs 79.2% respectively and analysis of three consecutive frames with ‘line border features’ showed accuracy of 85.1%. Exploration of social interactions between cows was addressed and the proposed seven-point cow shape model was investigated for this purpose. The proposed seven-point model may act as a crucial component towards building a robust system for automatic detection of social interactions in areas of interest.
|
|
8. |
|
|
9. |
|
|
10. |
|
|