This paper designs a low-cost pedestrian volume investigator orchestrating Raspberry Pi nodes, wireless network connectivity, and machine learning techniques. Under the control of a coordinator, 3 sensors capture the distance to the closest object in the target space for both learning and estimation. To obtain learning patterns, a human operator initiates a data acquisition transaction and records the number of objects he or she observes. With the set of learning patterns, each of which

more »

... of 3 distance measurements and the number of objects, we build a 3-layer artificial neural network model with 3 inputs, 20 hidden nodes, and 1 output. Next, the investigator periodically collects the sensor readings and estimates the number of objects. The simulation study shows the error size hardly exceeds 1 object until the number of objects is 8, indicating that the proposed scheme, as a new Raspberry Pi application, can economically trace thenumber of objects with reasonable accuracy and flexibility.