| At present,animal husbandry enterprises not only need to improve the production management level,but also deal with various challenges caused by the epidemic.The intelligent transformation and upgrading of animal husbandry has become the main direction of enterprise development.Promoting the application of artificial intelligence technology in animal husbandry,which can realize the development of animal husbandry.In the process of cows breeding,accurate identification of the estrus period of cows can make them timely pregnant,shorten the calving interval,improve the milk yield,and then improve the economic benefits of cows breeding.However,the traditional estrus identification of cows mainly depends on the experience of managers,it is not only time-consuming but also easy to cause inaccurate identification.Cows in estrus period have some special characteristics,such as excitement and increased exercise.So we can predict estrus period through steps counting.However,in the working process of accelerometer,it needs to continuously collect the acceleration to count steps,which leads to the processor running at all times,resulting in high overall power consumption.In addition,some cows have recessive estrus,which means less exercise.For such cases,the estrus period of cows can not be accurately identified by accelerometer.In order to solve the two problems that the accelerometer needs to run continuously and how to construct an accurate estrus prediction model,a set of low power earmarkers and system for estrus prediction is designed.The earmarkers realize the data collection function.The collected data is sent to the Bluetooth gateway through the BLE.After receiving the data,the Bluetooth gateway sends it to the trunk gateway through Lo Ra.The trunk gateway sends the data to the IOT platform through Wi-Fi,and finally the IOT platform realizes the data processing function.Firstly,due to the accelerometer needs to run continuously,which leads to high overall power consumption,an intermittent steps counting model is proposed.The earmarkers work intermittently to collect data and the cloud server trains the model to predict the steps.To further reduce power consumption,an intermittent steps counting model based on dynamic duty cycle is proposed.It can dynamically adjust the working mode according to the state of cows.Secondly,it is difficult to predict recessive estrus cows by accelerometer.An accurate estrus prediction model based on steps and body temperature is proposed.Based on the relationship between steps and estrus peroid,this thesis generates a reasonable data set from the perspective of statistics,and the data set is preprocessed to extract the feature vector.The accuracy of estrus prediction is evaluated by different classification models.In the intermittent steps counting model,the absolute time slot of the working cycle is set as 5s,the average working current is 0.76 m A and the average current is 7.7u A when the system enters sleep mode.When the intermittent steps counting model is used for experimental verification,the prediction results of 50%,70% and 90% duty cycle are compared.The root mean square error between the actual steps and predict steps is as follows:decision tree>random forest>KNN>SVM,according to the power efficiency ratio,the best model is SVM.This thesis designs a dynamic duty cycle mode,which can switch between50%,70% and 90% duty cycle according to the state of cows.In the accurate estrus prediction model based on steps and body temperature,the data set is generated according to the behavior of cows.In different machine learning models,The accuracy of cestrus prediction model is KNN(99.32%),SVM(99.42%),decision tree(99.51%),random forest(99.51%)and LVQ(99.06%),respectively.The best estrus prediction model is random forest. |
PDF Full Text Request