Study On Detecting Dead Birds In Modern Chicken Farm Based On SVM

With the information processing, artificial intelligence and image processing technologies becoming more sophisticated, the traditional ways of livestock farming has been changed a lot. In modern poultry farm, the supply of feed and water, the collection of eggs and chicken manure, the control of temperature and wind speed, etc, all have been automated. But for dead chickens, still using manual inspections and checks. Manual detection dead chickens has a low real-time, could not clear the dead chickens in timel. In addition, the efficiency of workers is not high, and prolonged the time staying in the hen house in a poor environment, health will be affected. For the shortage of manual detection dead chickens, this paper proposed a detection way for dead checkens which based on support vector machine and machine vision technology.Machine vision technology has advantages of no contact with detected objects, no time limits, high distinguish precision and speed and so on. Using machine vision technology to build this dead chickens testing machine vision system to realize the detection dead chickens automaticly. The system consists of hardware and software, the former is the base of detection system, which mainly accomplish the image acquisition; the later is the core and key of detection system, which mainly fulfill the image processing, feature extraction, image recognition and results determined.In the part of hardware system, explained how to choose a suitable CCD camera, lens, image acquisition card and other key components in the machine vision system, Built hardware test platform including PC, acquisition card and camera. In the part of software, Visual C + + 6.0 as the development platform, the software system consisted of image acquisition module, image processing module, image recognition module and information management module these four main modules.In the part of detection algorithm, using a theory based on statistical - support vector machine(SVM) algorithm to detect dead chickens. Firstly, according to the changes of central region of cockscomb in the picture, logic and operation is used to remove the part of live chickens. Secondly, in order to distinguish accurately the dead birds exist or not in the picture processed, the perimeter, area, eccentricity, complexity and roundness of the cockscomb are extracted out as the variables, the changes of those variables are defined as the feature vectors, training samples of the above feature vectors are used to train SVM classifier. Selecting LS-SVM classifier, and using grid search method to optimize the kernel width and punishment factor of Support Vector Machine and the SVM classifier for dead bird is designed finally.Experimental results show that: the proposed method detection accuracy is over 92%, and can detect dead chickens more timely and accurately. Besides, providing a theoretical and technical basis for further development of detection technology in poultry farm.