Image Processing Algorithm Research Oriented Bundled Bars Counting System

Counting the number of bundled steel bars is an essential part in the warehouse management and sales process of iron and steel enterprises. At present, in the iron and steel enterprises, the original manual method is commonly used to count the number of bundled steel bars. The original manual method is boring, heavy and inefficient, which may causes counting errors, even the phenomenon that the bar counter maliciously and falsely counts the number of bars. Therefore, the manual counting method has become a bottleneck which restricting the management automation of iron and steel enterprises. This subject is derived from a domestic iron and steel logistics company which has the actual demand in bundled steel bars counting automation, and develops the bundled steel bars automatic counting system by requirements of the company.In this paper, combining function and application of the bundled steel bars automatic counting system, a small portable counting system based on ARM9 embedded hardware platform is proposed, both of the hardware circuit and the counting software are designed and implemented. In the bundled steel bars automatic counting system, S3c2440A is used as the embedded microprocessor, Windows Embedded CE 6.0 is used as the embedded operating system, and based on this, this paper develops the counting software, which realized functions such as field image acquisition, real time image processing, and counting result showing,This paper proposes a set of method which is suitable for the subject, includes requirement image preprocessing, recognition and counting, and realized it in the embedded platform. This paper acquires bundled steel bars image by the digital image sensor OV9650, extracts region of interest by the polygon area selection method, preprocesses the region of interest to get the binary image of bar section, proposes a template matching based on connectivity and circularity detection to identify bars and show the counting result. Compared with existing template matching algorithms, there are two improvements as follows. Firstly, the algorithm probes the more likely match points by multiple templates, and only computes similarity on these match points, so that the computing time on these non-match points is reduced, the real time ability of the algorithm is improved. Secondly, the similarity measure method considering multiple features of the bar section, includes area, connectivity and circularity, it can effectively overcome the interference of residual pixels after the bar has been identified to the follow-up template matching, improve the counting precision and covering accuracy. In order to facilitate users to judge whether the counting result is correct, in addition to show the counting results with number, the counting software of this paper also identifies the position of the recognized bar with a red area.The experimental results show that the system can count the bundled bars effectively, and can satisfy the demand of the iron and steel enterprises on steel bars counting automation.