Research On The Multi-Sensor Data Fusion Algorithm For Pig Breeding Environment

Multi-sensor data fusion technology is a study of theory and method to process and use the multi-source integrated uncertain information comprehensively, which is capable to deal with the information in multi-grade, multi-aspect and multi-level process. With the development of agricultural information process and the data fusion technology, more and more attention have been paid to the application of using the multi-sensor data fusion technology in agricultural area.Pigs in the growth period requires high environmental condition, the deficiency of real-time monitoring to the pig breeding environment may lead to serious air pollution, thereby affecting the health of pigs. Using sensors to monitoring the environment in pig houses can help improve the informatization and scientific level of pig breeding of our province. However, the observation data from a single sensor may result deviation compare to the actual situation because of the instability of the sensor itself and the environmental noises. Applied the multi-sensor data fusion technology to the pig breeding environment monitoring can improve the monitoring accuracy and reduce the probability of misjudgment. The main contents of this paper include the following aspects:Introduced the main algorithms and the application researches domestic and foreign scholars has been achieved by using these algorithms in multi-sensor data fusion area. Compared the advantages, disadvantages and the scope of application of the common data fusion algorithms, highlights the methods that this paper used:the fuzzy set theory and the DS evidence theory.Proposed a two-level model of multi-sensor data fusion based on the background of pig breeding environment monitoring. The first layer of the model analysis the effectiveness of the sensors, using the improved optimal fusion set algorithm to preprocessing the data, excluding the unstable sensor observation data to prevent the impact on the accuracy of fusion. The second layer using the fuzzy membership function to calculate the basic probability assignment function of DS evidence theory, and calculate the degree of support among the evidence as weights allocated to the basic probability assignment function. Improved the Jousselme distance formula and introducing the improved formula to improve the conflict measure factor k in DS evidence theory.Monitored the temperature, humidity, ammonia concentration, carbon dioxide concentration in the experimental pig house, and use the proposed two level fusion model to fuse the monitoring data, compare to the result of classical algorithm, the proposed method reflected more obvious on cumulative probability. The experiment results achieve the desired goal, and demonstrate the improvements have a certain optimized effect. Make a decision to the experimental pig house and give a corresponding regulatory advice at last.