Research On Intelligent Inspection System Of Tile Production Line Based On Vision

Foshan is an important ceramic tile production town in China and even the whole world,among which most ceramic tile production enterprises have automatic production line,but the degree of intelligence of the production line is generally not high.With the aggravation of trade friction between China and the United States and the increasingly fierce competitive environment,transformation and upgrading have become the only way for these enterprises.This paper focuses on the demand of a famous ceramic tile manufacturer,which needs to solve two problems: first,the traditional photoelectric switch is used to calculate the output error of ceramic tile;Secondly,the roller conveyor on the production line is affected by the installation error,uneven load,vibration and other factors,and the roller way often runs out of the way,resulting in the loss of the brick body.At present,intelligent monitoring has been gradually applied in industry due to its advantages of independent judgment,but it has not stepped into the production line of ceramic tile.In this paper,combining with the actual production environment and using computer vision technology,the intelligent monitoring system of ceramic tile production line with application value is realized based on the detection and tracking of ceramic tile,which provides a new idea for the production statistics of ceramic tile production field and the detection of roll table deviation.Specific research contents are as follows:(1)aiming at the problem that the traditional target detection algorithm has poor detection effect when the target is similar to the background color and the target moves slowly,a target detection algorithm based on HSV color space is proposed.The target feature model was constructed with hue as the dominant and saturation as the auxiliary to distinguish the background.Meanwhile,the iterative clustering idea based on k-means clustering algorithm was used to provide the extraction method of component threshold value,and the component threshold value was accurately divided to distinguish the target from the background pixel so as to detect the target.It is proved that the algorithm can accurately extract the target region from the background with a lot of interference.(2)combining the characteristics of tile arrangement and morphological processing results,a blind segmentation algorithm of connected domain based on equipartition and area statistics is proposed,which can accurately extract the monomer tracked object.To eliminate the lag of the tracking algorithm,Improve the stability of tracking,using linear predictor to predict the location of the feature points in the next frame appears,and the use of multiple features fusion matching approach to tracking moving targets,including building a similarity measure function expression characteristics choose centroid distance and area difference,the result shows that to accurately target tracking object.(3)based on the tracking results after target detection,the "target chain" algorithm commonly used in the counting algorithm based on tracking is improved.The effect of short-term tracking failure is eliminated by judging the centroid displacement,and the reliability of output results is increased by combining multi-frame confirmation.In addition,the motion information of the ceramic tile can reflect the operation state of the conveying device,and the motion track can be fitted with the centroid coordinates obtained by tracking the ceramic tile,so as to judge whether the conveying behavior of the roller table is normal or not.