Research On Industrial Explosive Packaging Process On-line Detection Technology And Its Application

With the rapid development of economic construction and expanding production scale of the industrial explosives, civil explosive industry requires a production safety technology to be constantly improved, and the automatic and intelligence of the equipment gradually increase. Packing process is an important process of explosive production; the production process has been the realization of mechanization and automation, but equipment needs to be improved its safety and reliability as well as intelligent levels.Aiming at production process information deficiency and an explosive cartridge gesture which influence normal operation on an industrial explosive product line. This article research and development of an industrial explosive packaging process on-line detection system based on machine vision.Firstly, based on a production process of industrial explosives packaging line, specific needs of detecting task and machine vision system are indentified. According to the requirements, the machine vision system lighting solutions and imaging equipment selection are proposed. So the machine vision system is applied to an industrial explosive production field.Secondly, images of industrial explosive from the production process are analyzed, and the technology of image preprocessing is described. In order to simplify a cartridge description method, the cartridge is described for the linear model according to the shape of cartridge characteristic. Research on segmentation algorithm for cartridge image, an algorithm that is proposed to calculate the threshold, and at the same time the image is classified. The algorithm is to prevent possible error segmentation.Addressing at an abnormal transmission gesture of cartridge may threaten the explosive production line to operate, recognition of cartridge transmission gesture and diagnosis method are presented. An edge detection method and a rectangular approximation method are proposed to extract a line model from a segmented image which is based on the image segmentation. According to the gesture information in the image, cartridge transmission gesture is diagnosed by rules predefined. The experiment has shown that the recognition of the rectangular approximation method better and recognition rate can reach more than92.5%.For obtaining explosive information in the packaging process, a method base on image processing and support vector machine (SVM) to classify cartridge image foreground object is proposed. A gray-level distribution feature and shape feature of the image is defined as sampling data. A support vector machine is as the classification algorithm to classify the data. The parameter optimization of support vector machine is processed by a grid search method. The feature which is as input to train the SVM classification model and finally the classification model after training is treated as a classifier. The simulation result has shown that the selected features were fused by the classification algorithm. Compared with the single feature and BP neural network classification algorithm, the method has better identification rate, and the training time is less than BP neural network. The method can be applied to the production line to acquire image information.Finally, Based on the above approach and Microsoft Visual Studio2008as well as OpenCV function library, the on-line detecting system of industrial explosive packing process has been developed. The proposed system can quickly and accurately detect some information of cartridge production line and run steady. It can monitor the production line of industrial explosive.