Reflection tomography laser radar image segmentation method based on target area local enhancement

Abstract

The invention provides a reflection tomography laser radar image segmentation method based on target area local enhancement. The method comprises the following steps: inputting an original image generated by a reflection tomography laser radar and obtained through filtering back projection reconstruction; removing ring artifacts in the original image through a two-dimensional low-pass filter; further performing mean filtering processing to filter noise points, and performing initial segmentation on the image after noise point filtering by using an iterative threshold method to obtain a segmented image; target region extraction is carried out on the segmented image according to the criterion that the area of the circumscribed convex polygon is minimum, and a closed target region of the circumscribed convex polygon is extracted; performing image filling on the extracted target region by adopting an image filling algorithm to obtain a target saliency map; fusing the target saliency map with the original image to obtain a fused image with an enhanced target area; and solving an optimal threshold value of the enhanced target area image by adopting an iterative threshold value method, and then segmenting the whole fused image by using the optimal threshold value.

Description

technical field [0001] The invention belongs to the technical field of laser radar, and in particular relates to a reflection tomography laser radar image segmentation method based on local enhancement of a target area. Background technique [0002] Laser reflection tomography imaging is a new type of lidar system that combines long-range and high-resolution imaging. This technology was first proposed by Parker J K in 1988 (Parker J K, Craig E B, Klick D I, etal.Reflective tomography:images from rangeresolved laser radar measurements[J].Appl Opt,1988,27(13):2642-2643). Its working principle is to detect multiple angles of the target through lidar and collect echo signals, obtain the depth information of the target at multiple angles, and use the imaging algorithm to calculate and reconstruct the tomographic profile image of the target. However, in the actual work process, when using common imaging algorithms such as filtered back-projection algorithm to calculate the tomogr...

AI Technical Summary

Problems solved by technology

[0007] From the perspective of threshold segmentation images, the above three traditional threshold segmentation methods all have good segmentation effects on stripe artifacts and noise points, but ring artifacts and the gray value are close to the target image, so the traditional threshold segmentation method Difficult to separate it from the target

In addition, the three threshold segmentation methods all have different degrees of over-segmentation, and some target information is missing while removing artifacts and noise, especially the maximum entropy algorithm has the most missing target information.

[0008] The gray histogram of the original image is as follows image 3 As shown, it can be seen that there is only one obvious peak in the figure, which is caused by the background of the image. Select the gray histogram with a gray value range of 100-140, and mark the above-mentioned Ostu algorithm, iterative threshold method and maximum entropy algorithm. The segmentation threshold value, the segmentation threshold value selected by the traditional threshold segmentation method is usually high, prone to over-segmentation

However, because the gray value of the ring artifact is close to the target image, even if a higher segmentation threshold is selected, it is difficult to separate it from the target image by simple threshold segmentation.

Images