The Human Body Identify Method Study At Thecompleted Infrared Background

Nowadays, there are many kinds usage for the human body detection apply to the infrared camera. The auto shooting to enemy solder is one of usage at my documents.The principle part of this system is running in the following: to collect infrared information by in-active infrared camera as well as knowledge of both picture treatment and pattern detection; to analyze characteristic of the picture; to obviate the noise interruption by median filtering; to classify the gray level of picture by rank graph balance automatically according to the small gray difference for infrared picture. The body edge is not smooth, for a lot of burrs are there when we treat picture by above-mentioned method, besides, there still exist some noises on picture, which will effect negatively on the later treatment. Therefore we use erode and enlarge method to improve the picture; to analysis both advantage and disadvantage of normal edge detect mathematic formula; we select Prewitt mathematic formula based on actual picture characteristic; we apply average gray threshold for 5 picture gray threshold calculation in order to divide the picture.The picture treatment above will offer the next human body pattern detection. Among Normal human body pattern detection method, some are based on moving character, some on form information, some on walking model &etc. I put forward detection of the human body based on edge track theory. Chain code is one of important concept in picture treatment, and it can show 8 directions around the center pixel. The summation of chain code can represent the slope of the edge curve, and the difference of chain code can represent the curvature of the edge curve.Method from author here runs like this: to detect the human head firstly, the human shoulder secondly, and the human's arm and leg lastly, all of them are according to human body's characteristic.The bend direction of human head can be shown by the chain difference result, i.e. negative or positive, then whether human head is on the picture could be detected. During the detection process, successfully detecting half human head is enough to detect further on the human shoulder according to the coordinate got already. One segment of the human shoulder contour line is straight line, and the curvature at the out corner of shoulder takes maximum figure, based on which human shoulder could be detected. By analogy, detection of human arm and leg could also be done when half of human shoulder is detected.The edge of human body in picture must be smooth and continuous if we take this kind of picture treatment. We bring forward our own resolving concept for the breakdown of the edge, which we can scan the next high gray point from the break point as a starting point, till the next high gray point as the continuous curve line. We set a definite threshold and permit a definite failure points we have during the human body detection according to the actual status.The human body detection in this document is both important and difficult. The simulation verified that it is feasible to apply this pattern detection. We select a simple and efficient mathematic method, and by VC++ program we accomplish the simulation of human body pattern detection, whose result can meet the study target, whereupon one human body pattern detection method is put forward. This method can be applied into the future study activity.