Closed Space Dirt Recognition And Image Processing

Enrichment of boiler and piping inside dirt directly affect the life of the boiler, pipes, reducing the efficiency of the use of inside space. At this stage, it’s difficult to clean the boilers and piping inside. Firstly, the space waiting for being cleaned is mostly enclosed or small confined space, which makes the cleaning staff difficult to enter such space to do work. Secondly, manual cleaning process can only remove dirt of the wall surfaces, but can’t remove stubborn scale, which can’t reach the purpose of thorough cleaning. Based on the above situation, a dirt detection system being able to enter such enclosed and small confined space to do cleaning job is designed in the research, and under the enclosed non-regularization environment with complex internal structure, it can complete the inspection tasks of the inner wall of enclosed space. Moreover, image acquisition and processing is the core to achieve the above functions.At first, the hazards of dirt on the boiler, pipes and other confined spaces are proposed in the paper, and the importance of cleaning the dirt is cleared. Then the problems need to be resolved in the decontamination process are analyzed, in which the problems are dirt identification and distinguish. And then, the characteristics of enclosed space are described in detail, and the choice of the light source is analyzed. At last, the proper lighting, the right lighting way and the best color of the light source are selected to irradiate the object considering the characteristics of light, the lighting way and the effect of different color light sources on the images. The methods for determining the dirt inside the confined space are divided into the following areas:① Method of judging the presence or absence of the uniform dirt can be divided into two steps, firstly, making the collected images organic segmented, based on the set threshold, judging the uniformity of dirt through comparing the difference between the divided images. If the image contrast is within the allowable range inside, it is considered uniform. And the uniform image containing dirt or not is distinguished by comparing the images between inner wall of dirt and the initially pre-stored.② It’s need to make pre-processing before recognition the images of non-uniform dirt. The corresponding optimization program is proposed based on the in-depth analysis of a series of problems including. The key of pre-processing is to select the illumination source, the use of inappropriate sources may bring a lot of negative impact to the imaging system, increasing the difficulty of post-processing. Using white LED lights for illumination, when making image processing, the effect of algorithms to identify dirt is better, so a white light source is used in the paper.③ The non-uniform dirt and inherent parts after image pre-processing are distinguished through the neural network method, distinguishing the space inherent parts and dirt. Firstly, collecting the image features of non-uniform dirt and inherent parts, in which these features include the perimeters, areas and colors of the dirt and parts. Taking the image features of multiple samples of dirt and parts as input, and then based on gray neural network, the algorithm program of image acquisition and analysis is designed. Using MATLAB to optimize the neural network algorithm, re-establishing the mathematical model of image recognition, and investigating it through simulation and experiment.