Saliency-based road target extraction method in night-vision infrared images

Abstract

The invention discloses a saliency-based road target extraction method in night-vision infrared images and belongs to the field of machine vision. The saliency-based road target extraction method specifically includes: performing crude extraction of salient regions on the night-vision infrared images by applying a GBVS (graph-based visual saliency) model; performing re-extraction of the salient regions by applying a supercomplex number frequency domain based spectrum scale space method on the basis of global features; performing fusion of global and partial clues on saliency images ZZ by applying the saliency extraction method based on fusion of the global and partial clues to acquire final saliency images. With the method, saliency on time domain and frequency domain, global contrast and local contrast, low level features and high level features is fused, energy emission is considered to be a salient factor, processing objects are not normal natural images but infrared images, and the original natural images are extracted on the basis of the salient regions of the energy emission by processing the infrared images; extraction effect of the obtained salient regions is more accurate and comprehensive, and outline of salient region targets are clearer.

Description

technical field [0001] The invention belongs to the field of machine vision, in particular to a method for extracting road targets based on saliency in night vision infrared images. Background technique [0002] The visual saliency mechanism can be used to extract the salient areas that have a significant impact on driving safety in complex road scenes, laying the foundation for the post-processing of computer vision, improving the effect and efficiency of post-processing of computer vision, and facing complex road conditions. For road scenes, if all the content is processed by computer vision, the amount of data processed by the computer will be very large, and the processing efficiency will be very low. Therefore, researchers are inspired by the visual attention mechanism in the human visual system and hope that Machine vision, like the human visual system, first screens the original input data, thereby reducing the amount of calculation for post-processing and improving i...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the deficiencies of the prior art, the present invention proposes a method for extracting road targets based on saliency in night vision infrared images, which solves the problem that the existing visual attention model has a low correct rate in target detection in night road scenes and is easy to detect. Affected by non-salient areas in natural scenes, easily affected by lights in night road images, and the outline of salient targets is not clear, etc.

Images