DESIGN AND IMPLEMENTATION OF REAL-TIME MONITORING SYSTEM FOR COW’S ESTRUS BASED ON STORM

The identification of cow’s estrus is one of important production activities in dairy farm.In tradition,the decision of cow’s estrus often relies on the breeder’s manual inspection and judgement.It’s time-consuming and inaccurate for large-scale dairy farms in current times especially harder at night.Moreover,that the period of estrus is too short makes it harder for staff member of dairy farms to forecast the estrus of cows.What’s worse,that the failure rate of artificial insemination has increased year by year and dairy production potential can not be effectively utilized greatly make the economic benefits of dairy farmers reduce much because missing the period of cow’s estrus and ineffective insemination often take place.Based on domestic and foreign research of monitoring and recording changes of cows’ body parameters with exercise and body temperature sensors,a variety of cow’s estrus information prediction model were put forward and much success was achieved.However,there are still some problems.For example,timeliness is poor and the accuracy needs to be improved and so on.In order to achieve accurate and rapid prediction of cow’s estrus,a real-time monitoring system for cow’s estrus was designed and implemented based on Storm which is a real-time streaming framework in the field of the big data.The system provides an effective tool for the prediction of cow’s estrus and also provides some guidance for the monitoring of other large animals.The main work and conclusions of this paper are as follows:(1)In order to obtain accurate and time-efficient cow’s data of body parameters,combined with the production environment of dairy farm,the programs of acquisition and transmission of cow’s data of body parameters were determined.High performance of pedometer called AfiTag II was directly selected to acquire cow’s data of body parameters,and then data obtained by pedometer was transmitted to the control room of dairy farm by two bridged TP-LINK TL-BS210 antenna.Thus,body parameters of cow like step,lying time,standing time and bout could be obtained regularly and automatically.(2)The effective body parameters for prediction of cow’s estrus were put forward.Based on the analysis of changes in body parameters of cows during the period of estrus,taking 2h as a single time slice and 6h as a significant sliding window,the number of steps of consecutive 3 slice units in the significant window of estrus s1,s2,s3,the cumulative standing time t1,cumulative bouts m and cumulative lying time t2 in W were finally selected to be the effective feature vectors for the prediction of cow’s estrus.The experimental results showed that the accuracy of prediction using the scheme mentioned above was higher than other selection schemes.This scheme could distinguish whether the cow is in the period of significant estrus period.(3)The SVM model and BP neural network model were constructed to predict cow’s estrus.And then,the model parameters were optimized.On the basis of constructing SVM and BP network prediction model respectively,kernel function,penalty factor and kernel function parameters of SVM model were optimally selected through experiment.The test data set concluded 240 groups of body parameters in estrus and 720 body parameters in non-estrus.The test results showed that the accuracy of the SVM prediction model using the RBF kernel function(penalty factor C=32,kernel function parameter g=3.81e-06)was 86.05%,the cross-validation recognition rate of the model was 98.69%,and it had high estrus prediction performance.(4)The cow’s estrus real-time monitoring system and information management system based on Storm was designed and implemented.In order to realize real-time and visual monitoring for estrus information of cows and to expand information management and decision-making of dairy farm in the future,the system was based on the B/S framework and was deployed with some services such as herd management service,pedometer management service,summary service,system functions management service,and log management service and so on.At the same time,real-time processing platform based on storm was set up.Combining the characteristics of processing platform of Storm,the prediction method of cow’s estrus was integrated into the real-time processing platform.And then,the result of estrus information was informed of the farm staff via SMS.(5)The entire hardware and software system was deployed and tested for performance in dairy farm.The results of tests showed that the average delay of the system was within 2s,the average accuracy rate was over 98.50%,the prediction accuracy of cow’s estrus was 93.33%,the misjudgment rate was 2.40%and the cycle of estrus was shortened to 2h,which meet the needs of practical production and provide effective technical support for timely breeding of cows in dairy farm.