Data Fusion Method for Temperature Monitoring in Livestock and Poultry Housing Based on Improved Support Function

Abstract:

Insufficient accuracy in dynamic temperature field reconstruction significantly constrains the responsiveness and control efficacy of precision feeding decision systems. To address this critical limitation, a novel dynamic data fusion algorithm was proposed based on an improved support function (Fast-NF). The algorithm constructed a spatio-temporal weight optimization model for sensor data by coupling dynamic time warping (FastDTW) with a multi-layer polynomial decay mechanism. This approach effectively overcame the technical limitations of traditional methods, particularly concerning computational efficiency and compensation for missing data. Experimental results demonstrated that compared with conventional Gaussian-based methods, the root mean square error (RMSE) of the fusion temperature was reduced from 0.436 7℃ to 0.387 5℃, a decrease of 11.3%. The window calculation time was shortened from 14.568 8 s of the standard DTW-NF algorithm to 5.839 4 s, and the efficiency was improved by 59.9%. Consequently, the developed dynamic temperature field reconstruction technology, leveraging this method, successfully achieved dimensionality elevation from discrete monitoring points to a continuous field domain. This advancement can provide robust support for precision feeding decision systems, offering a significant improvement in reconstruction accuracy and computational efficiency for practical agricultural applications. The core innovation lied in the synergistic Fast-NF mechanism integrating FastDTW and decay weighting.