Local Adaptive Threshold Segmentation Method for Wear Particle Recognition in Online Ferrographic Image

Abstract

The local adaptive threshold segmentation method for online ferrographic image wear particle identification firstly converts the online ferrographic transmitted light image into a grayscale image and a Laplacian image after preprocessing. The image is calculated by its integral image to reduce the time complexity of subsequent processing. Secondly, combining the distribution information of the two, an adaptive model of the scale factor t is established, so as to obtain an adaptive model of the local threshold Threshold. Finally, compare each pixel value of the grayscale image with its corresponding threshold pixel by pixel, and distinguish it as the target or the background, and complete the segmentation of the online ferrographic image. Based on the idea of ​​local threshold segmentation and combined with the characteristics of wear grain images, the present invention establishes a complete automatic segmentation mechanism. The problem of automatic segmentation of background.

Description

technical field [0001] The invention relates to the field of fault diagnosis and state monitoring of mechanical equipment, in particular to an online ferrographic image digitized automatic recognition technology, in particular to a local adaptive threshold segmentation method for online ferrographic image wear particle recognition. Background technique [0002] Condition monitoring and fault diagnosis technology is an important guarantee for the reliable operation of modern major equipment. Ferrography analysis technology, as a means of tribological state monitoring and diagnosis, can extract the wear state information of mechanical equipment by analyzing the information such as the number, shape and size of equipment wear particles carried by lubricating oil, and can diagnose wear faults and formulate corresponding visual Environmental protection strategies play an important role. Off-line ferrography analysis technology has formed a very mature analysis method and is wide...

AI Technical Summary

Problems solved by technology

This method can accurately extract a single target wear particle from the ferrographic wear particle image with a single background color and a "single peak" in the gray histogram, while the gray histogram of the online ferrographic transmitted light image is mostly double-peaked, and the background will appear dark side

Therefore, the object of the above research and analysis is limited to offline ferrographic images, and cannot be fully applied to online ferrographic images.

[0004] For the segmentation of online ferrographic images, Qiu Huipeng et al. used the Otsu maximum inter-class variance method, iterative threshold segmentation, etc. to automatically obtain thresholds in the "Research on Image Information Extraction Technology of Image-type Online Ferrography", and there are still errors in segmentation. , a background difference algorithm based on ferrographic image features is proposed. Although the result of the segmented wear particle area is relatively accurate, this method needs to obtain a reference image without wear particles in advance. Due to the influence of real-time and variable working conditions, it cannot achieve ideal results. Effect

In addition, Tao Hui et al. combined edge detection with threshold segmentation in "Design and Application of Improved Otsu Based on Ant Colony Algorithm in Online Ferrographic Image Analysis", and improved Otsu method based on Ant Colony Algorithm to obtain the best threshold and perform binary However, the global thresholding method is greatly affected by the local singularity of the image. Under factors such as noise and uneven illumination, the gray level of the background or the object changes, and it will be difficult to segment the entire image with a unified threshold. high false positive rate

Images