Abstract: With the rapid development of smart agriculture, agricultural sensors have played an important role in promoting the iterative upgrading of agricultural science and technology and the transformation of agricultural production methods. The agricultural environment monitoring objects are complex and diverse, but the sensor types are few and the precision is low, so the monitoring data obtained is large and redundant, resulting in poor reliability of information perception. In order to improve the problem of low accuracy and poor reliability of single-sensor measurement data, In this paper, an improved radial basis function neural network (RBFNN) and Chernobyl disaster optimizer (ICDO) multi-sensor data fusion algorithm is proposed. Firstly, an improved Chernobyl catastrophe optimization algorithm is proposed to optimize the neural network model. By introducing the good point set theory, the initial population quality of the CDO algorithm is improved, and the accuracy and speed of the algorithm are improved. By adding the adaptive Laplacian crossover operator, the search performance of the algorithm is enhanced, the algorithm has better adaptive behavior, and the convergence speed of the algorithm is accelerated, and then, the individual learning strategy and differential evolution strategy are used to redefine the location update equation, which balances the local and global exploration capabilities of the algorithm. Secondly, the RBF neural network model is optimized by ICDO to improve the stability of the model. Finally, the nonlinear mapping capability of RBF neural network model is used to realize the multi-sensor data fusion method, which improves the data fusion accuracy. In order to verify this algorithm, three experiments were conducted in this study. The first one is the verification of ICDO algorithm, which has a large improvement in solution accuracy and optimization stability compared with particle swarm optimization (PSO), gray wolf optimization (GWO), firefly algorithm (FA), dung beetle optimizer (DBO), and subtraction average based optimizer (SABO). The second one is the simulation of atmospheric environment quality judgment, which was studied on the atmospheric data collected outside the South Subtropical Botanical Garden in Mazhang District, Zhanjiang City, Guangdong Province, China, from September 1, 2022 to September 30, 2023, and the improved model was validated. The goodness of fit of the atmospheric environmental quality prediction reaches 0.999, the mean square error is as low as 0.348, and the mean absolute percentage error is reduced to 0.729%. The third one is the greenhouse environmental classification, which classifies the data collected in the greenhouses of the South Asian Tropical Botanical Garden into environmental classification. The accuracy rate of greenhouse environment classification is 99.21%, and the accuracy rate is 99.91%. The results of research prove that this fusion algorithm can get high precision fusion values for both indoor and outdoor environment data, with good adaptability and accuracy. This algorithm not only solves the key problems in agricultural sensor data fusion, but also provides solid technical support for agricultural intelligence, demonstrating its great potential and development prospects in the field of intelligent agriculture.