Pig Behavior Identification Based On Multisensor Fusion

In recent years,China's animal husbandry industry carry out in-depth cooperation with the various countries.Pig aquaculture gradually make great progress towards the intensive,automatic direction.At the same time,the domestic demand for meat products in the world.But the scale of the expansion,scientific and automated farming technology still lags behind the developed countries.Part of the technology of pig breeding still relies on traditional experience,ignoring the rule of pigs' behavior,there are problems such as improper design of the material slot,waste of feed and water resources,pigs' death because of untimely treatment,and so on,which bring many economic losses to enterprises,restricting the development of scientific automation of pig farming enterprises.In view of the backwardness of scientific automatic technology of animal husbandry in China,and the problem of grasping the rule of pig behavior,this paper designs a kind of pig behavior identification based on multi-sensor fusion.It is observed that the behavior of pigs is often accompanied by sound and movement.So the two behaviors are collected by sensors,and the data of sound behavior and motion behavior are sent to data processing center(Contains microcontroller STM32 and DSP5416)by wireless network.The data of two kinds of behaviors are calculated by data processing center.According to the characteristics of sound and motion,different eigenvector is used to synthesize the behavior and motion eigenvector,and the result of behavior identification is obtained.Then the identification results are sent to the monitoring platform in computer monitoring room.Monitoring platform shows pig posture and marks abnormal behavior of pigs.At the same time,the data is stored in the database to facilitate the breeder to query historical data.The main work of this paper is as follows:(1)Design of pig behavior identification scheme.This paper analyzes the demand of pig behavior identification in animal husbandry,introduces the layout status of Pigsty in the experimental site designed in this article.The pig distribution characteristics determines the behavioral identification scheme.By observing,the behavior of pig only occurs through movement and sound behavior.We complete this design from two aspects of sound and motion,design hardware circuit and choose data transmission mode,finally design the fusion scheme according to the behavior characteristics.(2)Design of sound identification system.Set up the voice data processing center DSP the smallest system,according to the scene collection sound characteristic,carries on the filtering,enlarges and so on.Then the eigenvector of the preprocessing sound data is computed and the eigenvector is simulated.The feasibility of the recognition scheme is validated.According to the abnormal sound frequency of the pig set by LM567,the TDOA is positioned in the pig with abnormal sound.The location of the pig position information is sent to the data processing center STM32.(3)Design of motion identification system.According to the distribution of pig and the characteristics of movement behavior,this paper designs the collection circuit and data transmission system of pig motion behavior.Data Processing Center Micro-Controller STM32 calculates the characteristic vectors such as acceleration,correlation coefficient and so on.Then the characteristic vector is simulated to prove the feasibility of this scheme.(4)The application of feature level fusion system for sound and motion behavior based on SVM(Support vector Machine).According to the characteristic vector of the pig behavior which is composed of the sound characteristic vector and the motion characteristic vector,the characteristic fusion is accomplished by calculating the kernel function,and nine kinds of behavior is identified(Standing,left lying,right lying,sitting,lying,eating,drinking,coughing,starving).The identification results are sent to the computer monitoring room for display and the identification results are stored in the database.According to the behavior regularity of pig,the behavior evaluation system of the behavior of time domain and frequency domain was used to evaluate the behavior status of pig.The evaluation system is labeled as a pig with abnormal behavior and the location information is displayed in the interface.(5)Design of computer monitoring platform.The computer monitoring platform shows the results of the fusion sound and the movement behavior,and marks the pig with abnormal behavior.Using SQL Server database to store the information of pig behavior,time and position.The experimental results show that the pig behavior identification system based on multi-sensor fusion can identify nine kinds of behaviors with high accuracy,and can locate the pig with abnormal behavior.Then display it in the computer monitoring platform.At the same time,the health evaluation of pig behavior,so that the keepers timely abreasting of pig's health conditions,the liberation of the human resources of aquaculture enterprises.The result of behavioral identification is stored in database,which provides data support for establishing daily behavior model.