Study Of People-Counting System In Marketplace Based-on Neural Network

As the entry of many foreign retail enterprises, the competition is fierce to retail day by day. Customer-counting is very important, because it is a foundation of retailing and it is proportional to sales amount. The traditional counting method artificially can't offer the real-time flow of customer data, while the rate of accuracy, which of simple counting with photoelectric transducer of infrared ray, is low. So it is an important route to improve the rate of accuracy of infrared-counting with pattern recognition technology of artificial intelligence.This paper designs a new people-counting system based-on neural network, According to the research approach of pattern recognition system and the Characteristic of the data, which photoelectric transducer of infrared ray gets. This system includes four photoelectric transducers of infrared ray, which are fixed in the entry of Market, and the height to the ankle. When the customer is through counting the area, it will produce a pattern. And the pattern is dealt with by the neural network of intelligence. It counts the number of customers and saves it. The groundwork is as follows:This paper chooses four photoelectric transducers of infrared ray, in order to distinguish customers which enter at the same time. And the initial data is preprocessed for strengthening the validity of data.Data of customers is continuous space-time sequence. According to this characteristic, this paper has proposed an adapted method of data segmentation. And the experiment proves this kind of method is effective.This paper has proposed an extraction method of feature parameter based on pulse-pulse sequence. And the experiment indicates this kind of method is effective.This paper chooses resilient Black-Propagations. And it learns a lot of samples and used to count the number of customs. The experiment indicates this system is valid. It can distinguish customers between whom there are some spaces, and the rate of accuracy attaches to100. And for customers who entry at the same time, it also has a high rate of accuracy. Besides, the experiment proves the model of customer-counting is correct.