Research On Smart Perception And Intelligent Analysis Of Agricultural Data

With the development of the Internet and Internet of things(IoT)technology,there are large amount of data related to before,during and after production accumulated in agriculture,resulting a variety of agricultural information service system emerge in endlessly.While for agriculture new production and operation staff,the cost and barriers of agricultural information service are high,the rural information service system is relatively scarce,and big data coexists with information island phenomenon.Agricultural information service is generally geared toward specific area,specific application,and specific data resource,making it harder to develop and utilize the data.Due to the lack of effective analysis,the conversion of data resources to useful information is inefficient,thus bringing about the contradiction between the enrich of data resource in agriculture and the scarce agriculture information service.Therefore,this paper studies the agricultural IoT and the collection and extraction technology of Internet data to realize the intelligent perception of agricultural data;For the characteristics of multi-sourced heterogeneous agricultural data and the specific business requirements in different agricultural scenarios,this paper studies the intelligent analysis and mining methods of agricultural data to realize the transformation from data to information;By establishing user interest model to study user clustering and personalized recommendation algorithm,this paper can realize the personalized precision agriculture information service guided by user interest.The research is of great and practical significance to improving the level and quality of agricultural information service and promoting the process of agricultural modernization in China.In recent years,there has actually been a lot of research done on big data analysis and personalized recommendation.However,there are still problems such as unstable data sources,low data quality,low decision accuracy of information mining and weak pertinacity of service push.In order to make better use of information technology services for farmers to increase agricultural production and operation efficiency,this study discusses the following aspects:agricultural Io.T,Internet data intellisense,the intelligence of multi-source heterogeneous data mining oriented agriculture analysis method based on user interest model and typical application scenario and several levels of personalized information push on agricultural data perception and intelligent analysis of key technologies in-depth study and discussion.The main research contents in this paper are as follows:(1)There are some problems in the process of IoT data perception collection under the complex agricultural environment,such as low quality,poor stability and poor timeliness.In this paper,an efficient data collection method of agricultural IoT based on data association analysis is proposed.To overcome the channel link instability problem in complex agricultural environment,the opportunistic routing is adopted to improve data transmission reliability and network data throughput.According to the characteristics of agricultural application scenarios,focusing on the bandwidth and resource constraints of the network,the weight of opportunistic routing relay selection and coordination is optimized from the perspective of node data correlation analysis and energy consumption perception,so as to realize the efficient perception collection of IoT data.In the aspect of Internet data collection,aiming at the problems such as complicated data and missing information extraction in the field of Internet agriculture,this paper studies the intelligent extraction method of Internet information in the vertical field of agriculture.Through fusing a variety of feature of agriculture Web pages,combined with rule learning algorithm of machine learning,an Web information extraction method based on feature and self learning is proposed.The proposed extraction method can effectively identify the body content part of the agricultural information website without human intervention,extract the agricultural information entity data,and adapt to Web page structure changes.(2)With the rapid development of the agricultural IoT and Internet technology,a large number of agricultural data resources have been accumulated and formed.How to effectively excavate the useful information is a problem that needs to be solved.In view of the characteristics of agricultural big data such as multi-source,heterogeneous and noisy,this paper studies data preprocessing methods to improve the quality of big data through cleaning,filtering,and integration,further ensuring the availability of data.On the basis of data preprocessing,an improved Apriori method based on frequent sets is proposed to analyze association rules among multi-dimensional big data,so as to explore the rules and values of massive heterogeneous agricultural data and provide the basis for realizing efficient information services in the agricultural field.For the problem of information asymmetry between supply and demand for agricultural products prices,based on the extracted agricultural products market price information in Internet,the data correlation analysis regarding to time and space distribution of agricultural prices is performed.The data correlation analysis can find sensitive and abnormal volatility,and predict the certain trend of agricultural product price changes and the influence of other agricultural products price fluctuations,thus providing support to preliminary decision-making in agricultural production and management and marketing.(3)To solve the problems of early detection of animal breeding and disease,a cow target recognition method based on image entropy on the basis of data clustering and image recognition in agricultural IoT is proposed.Through calculating the minimum bounding box area in target object contour map,the proposed target recognition method can capture mount behaviors in real time.Combining the characteristics of the hoof and the back,the exercise amount of the identified cows is fused for 7 consecutive days to judge the abnormal behaviors that affect the healthy reproduction of the cows and improve the scientific degree of dairy farming management.(4)Agriculture information owning the characteristics of regional,timeliness and periodicity is diversified and classified,and user demand vary with these changes.How to implement accurate recommendations for mining results according to the user characteristics is a key link in the process of agricultural intelligent information service.To this end,this paper presents a personalized recommendation method based on user interest model.The recommendation method can obtain the observable characteristics(occupation,plant type,cultural level,geographic location,and etc),and extract recessive characteristics(personality preferences,planting experience,NongQing farming,and etc).Further,establish a context integrated user interest model,and realize the precision push according to the user demand based on the collaborative filtering recommendation method.