Weight estimation is a common practice throughout many industries, though it typically requires that the objects to be weighed remain motionless. More often than not, it is beneficial to allow objects to move freely through a process, so that time is not lost in stopping and rerouting the object to a weight sensor. This is the basis for achieving dynamic weighing, where the object to be measured continues to have motion relative to the weighing sensor. Typically, this has been achieved with signal processing techniques that produce favourable results with singular objects. The challenge is when multiple objects are grouped and moving together;​ that is, they are non-singulated and cannot be weighed separately. This work reports the development of an In-Motion Weight Sensor array, which is a new dynamic weighing system with a new real-time signal processing method for estimating the weight of multiple, non-singulated objects. The array system employs a recursive least squares estimation algorithm to combine weight sensor data and the locations of boxes that are travelling through the array to attribute fractions of a box’s load to the appropriate individual sensors. To demonstrate the performance of the proposed system, a full-scale experimental setup has been built and tested. Through statistical analysis of the weight estimates of a variety of groups of objects, it is shown that the system can produce results within 10% measurement error for the majority of non-singulated cases. It is most effective for non-rigid boxes that also fall within the mid-range for package size and weight, around 0.05m² and 1-3kg, respectively. Changes to the mechanical design can vastly improve performance accuracy and precision, and recommendations for these alterations are given in the conclusion.