Numerous radar warehouses equipped with Internet of Things (IoT) sensors have recorded a large number of environmental monitoring data. These valuable data can be used to test decision systems and help to modify them. When the measured data are applied in real scenarios, weak components always appear to interfere irregular event surveillance Lots of missing samples, however, exist in IoT data, leading to a serious disturbance in frequency domain which may probably veil the weak components. Traditional algorithms obtain very limited effects when there exist quantities of missing samples. Based on this, a new weak components extraction algorithm is proposed in this paper for real temperature data acquired by IoT sensors installed in the radar warehouses. This algorithm suppresses missing samples disturbance by minimizing the mean squared error of inverse-transformed data and original data in available samples. Its effect is better than any other interpolation method, particularly when missing samples exist consecutively. Some experimental results verify the validity of the proposed method.