Increasingly, detecting systems are facing the common problem of information processing owing to low SNR, low data rate, low resolution, and low information dimensions, and it is called the weak observation problem. This paper analyzes its origin and proposes the concept of information assembling by using the repetition and prediction properties of the object of interest. Then, the probability of the cloud inference method based on Bayesian theory is proposed to address a weak observation problem such as state estimation. Eventually several new requirements for sensor design, information processing, and system control are discussed, which are three crucial factors in information system design.