Research On Elevator Door Safety Detection Technology Based On Image Sensor

With the social and economic development,the modern high-rise buildings are constantly emerging.As an indispensable means of transportation in the high-rise buildings,the quantity and growth rate of the elevator are also increasing.Elevator technology is becoming more and more mature,but in recent years,accidents caused by elevator accidents are common.The proportion of elevator door area accidents is more than 80%,which indicates that there are some defects in traditional elevator door safety inspection technology.In order to make up for the shortcomings of the traditional elevator door safety detection methods such as the light screen and the safety touch plate in the aspects of lack of sensitivity,the existence of blind area and non prejudgment,the elevator door safety detection technology based on image processing technology is researched to improve the reliability and accuracy of elevator door safety detection.First,the accurate extraction of Region Of Interest(ROI)is studied.Insufficient binarization effect of the unimodal or nearly-unimodal peaks of the gray histogram for the traditional Otsu method.And the status quo of the elevator door image binarization is not ideal for various improved Otsu methods.Combined with the characteristics of elevator door gray histogram,a new Otsu improvement method based on trough mixing weighting factor is proposed on the basis of traditional Otsu method.The improved method considers both the image entropy of the target area and the frequency information of the neighborhood of the threshold in the binarization process.Make the best threshold of the Otsu method near the valley of the grayscale histogram and the segmented target area is closer to the elevator door ROI.At the same time,combining the gray value anddistribution characteristics of the elevator door pixels,it is proved that the ergodic range of the algorithm is reduced by inference and the operation efficiency of the algorithm is greatly improved.As compared with the segmentation results of traditional Otsu method,other valley-emphasis methods and neighborhood valley-emphasis methods,the image entropy of targeted region was decreased 0.08 to 0.68 by the improved algorithm.Additionally,calculation efficiency was improved by 2 to 6 times as compared with other valley-emphasis methods and neighborhood valley-emphasis methods under the same simulation environment.Finally,in the binarized image,the ROI of the elevator door is accurately extracted by using the connected region labeling method and the minimum external distance method.In order to realize the function of obstacle detection and elevator door identification in the extracted ROI,it is necessary to detect the edge of the elevator door and extract the effective judgment edge.So an improved Canny algorithm was proposed.The algorithm uses a improved bilateral filter based on the intensity similarity to smooth the denoising,and used the gradient template in eight direction to calculate the gradient within 5 × 5 neighborhood.The algorithm specified a new non-maximal suppression rule to refine the edge,and used Otsu method to select high and low thresholds adaptively.Simulation results indicate that the improved Canny algorithm was superior to the traditional Canny algorithm and some improved Canny algorithm in the application of edge detection in elevator door images,and the connectivity was improved by 57%-71%.Finally,the determination rule is set to accurately identify the obstacle and its position by using the relationship between the center distance between the judgment edges and the center position.