| Modern animal husbandry farming has the trend of large-scale and intensive farming.Therefore,timely and accurate identification of cows in heat can improve the conception rate of cows and increase the economic benefits of farms.At present,most farms mainly rely on experienced technicians to determine cows in heat based on experience and combined with rectal testing,which is not suitable for current large-scale,intensive farming needs.However,the estrus detection system built based on contact devices such as pedometers has a strong cow stress response and is not suitable for welfare farming needs.Therefore,this thesis proposed a non-contact cow estrus detection method based on thermal infrared image acquisition device.The main research contents are as follows:First,Realize the temperature and humidity information data acquisition and wireless transmission function of the thermal infrared image acquisition area based on STC89C52 microcontroller,DHT22 temperature and humidity acquisition module and ESP8266 WIFI module.Second,Analyzing the thermal infrared image features of cows in key parts,this thesis proposes a method for feature extraction based on Lab color space thermal infrared images of cows,and discusses the effects of different feature contours on the maximum,minimum and average temperatures of cows after feature extraction,and determines the best feature contours of the eye and vulva,with an accuracy rate of 81.70% in feature segmentationThird,Based on the body temperature change characteristics of cows in estrus,the Logistic model and SVM model were used to establish cows in estrus identification model.The experimental results showed that when the maximum temperature,minimum temperature and average temperature at the eye and vulva of cows were used as inputs,the detection rate of cows in estrus based on the Logistic model was 79.10%,and the detection rate of cows in estrus based on the SVM model was 78.46%.As well as when the maximum temperature at the eye and vulva of the cow was used as an input,the estrus detection rate increased,with 82.37% estrus detection rate for the Logistic model-based cow estrus model and 81.42% estrus detection rate for the SVM model-based cow estrus model. |
