Catenary image segmentation method based on SPCNN and minimum cross entropy

Abstract

The invention discloses a catenary image segmentation method based on a SPCNN and the minimum cross entropy. The catenary image segmentation method comprises the following steps that firstly a standard pulse coupled neural network PCNN is simplified so that a simplified pulse coupled neural network SPCNN model is obtained; then value assignment is performed on the parameters of the SPCNN model; then the number of times of iteration is determined by utilizing the principle of the minimum cross entropy; and finally binary segmentation is performed on catenary images by utilizing SPCNN model so that the segmented catenary images are obtained. The experiment result shows that the segmentation effect of the method is better than that of an OTSU method and an iterative method, and the method shows a great segmentation effect especially for segmentation of catenary masts, insulators, pole plates and other parts and components. The image segmentation effect measure index VOI (Variation of Information) and PRI (Probabilistic Rand Index) are also better than those of the OTSU method and the iterative method.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, in particular to a catenary image segmentation method based on SPCNN and minimum cross-entropy. Background technique [0002] At present, my country's high-speed railway is developing rapidly, and railway safety issues are becoming increasingly prominent. Catenary is the main structure of the railway line, and its working status has an important impact on whether the train can receive electricity stably. Therefore, it is necessary to detect the working state of the catenary to ensure that it is in good working condition. At present, the traditional manual fixed-point inspection and inspection vehicle inspection methods are difficult to meet the real-time and reliability requirements. With the rapid development of digital image processing technology, catenary detection technology based on image processing technology is a real non-contact detection method, which is not susceptibl...

AI Technical Summary

Problems solved by technology

[0005] In the image segmentation processing of the standard PCNN network, there are problems such as difficulty in computer model construction, large amount of calculation, slow running speed, many network parameters in the model, and difficulty in automatic selection.

Images