CN-121999167-A - Three-dimensional reconstruction method, recording medium and system based on front scanning sonar image

CN121999167ACN 121999167 ACN121999167 ACN 121999167ACN-121999167-A

Abstract

The invention belongs to the technical field of underwater acoustic imaging and three-dimensional reconstruction, and particularly relates to a three-dimensional reconstruction method based on a front scanning sonar image, which comprises the steps of obtaining a two-dimensional gray-scale image acquired by a sonar, carrying out feature analysis on the image by using semantic segmentation to obtain bright and dark feature channels of a highlight region and a shadow region, and extracting boundary features; and extracting the gray gradient and gray ratio in the highlight region, solving the high profile point of the obstacle by taking the front edge point and the shadow boundary point of the obstacle as constraints through nonlinear fitting, and finally fusing the characteristics of multiple array elements to realize three-dimensional morphology reconstruction. The invention can effectively inhibit underwater artifact and noise interference, realizes high-precision three-dimensional reconstruction, has the advantages of light model, low power consumption and the like, and is suitable for hidden detection of complex underwater environment and underwater mapping scene. The invention also provides a non-transient readable recording medium storing the program of the method and a system comprising the medium, and the program can be called by a processing circuit to execute the method.

Inventors

WANG HUIGANG
LIU KE
ZHAO WEINA
Lei can

Assignees

西北工业大学深圳研究院

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (8)

1. The three-dimensional reconstruction method based on the front scanning sonar image is characterized by comprising the following steps of: s1, acquiring a two-dimensional gray image acquired by ship front scanning sonar; S2, carrying out feature analysis on the two-dimensional gray level image through a semantic segmentation network to obtain a bright-dark feature channel comprising a highlight region and a shadow region, and extracting corresponding boundary features; s3, extracting gray gradient and gray ratio value in an unsupervised manner in a highlight area; s4, nonlinear fitting is carried out by utilizing the gray gradient, the gray ratio and the physical characteristics of sonar reflection, and height contour points of the obstacle along the fore-and-aft direction of the ship body are calculated; s5, fusing boundary features and height contour points of multiple array elements to achieve three-dimensional shape reconstruction of the obstacle.
2. The three-dimensional reconstruction method based on the front scanning sonar image according to claim 1, wherein the semantic segmentation model in the step S2 is a convolutional neural network, the original U-net network architecture is used as a basis, a cavity convolution kernel and residual connection are introduced into a network module, and a separation channel parallel structure is adopted.
3. The method of claim 2, further comprising the step of using dense connections within the network modules, with non-dense connections between network modules.
4. The three-dimensional reconstruction method based on the front scanning sonar image according to claim 3, further comprising the step of connecting different layers of the U-Net network architecture to perform feature fusion.
5. The three-dimensional reconstruction method based on the front scanning sonar image according to claim 4, wherein the physical characteristics of the sonar reflection comprise the obstacle distance parameters fed back by the sonar device under the constraint condition that the front edge points and the shadow boundary points of the obstacles are at the same moment.
6. The three-dimensional reconstruction method based on the front scanning sonar image according to claim 5, wherein the physical characteristics of the sonar reflection comprise two-dimensional gray-scale image reflection characteristics extracted by the semantic segmentation network under the constraint condition that the front edge point and the shadow boundary point of the obstacle are at the same moment by the same sonar array element.
7. A non-transitory readable recording medium storing one or more programs comprising a plurality of instructions, which when executed cause a processing circuit to perform a three-dimensional reconstruction method of any one of claims 1-6 based on a front-scan sonar image.
8. A three-dimensional reconstruction system based on a front-scan sonar image, comprising a processing circuit and a memory electrically coupled thereto, wherein the memory is configured to store at least one program, the program comprising a plurality of instructions, the processing circuit running the program to perform the three-dimensional reconstruction method based on a front-scan sonar image of any one of claims 1-6.

Description

Three-dimensional reconstruction method, recording medium and system based on front scanning sonar image Technical Field The invention belongs to the technical field of underwater acoustic imaging and three-dimensional reconstruction, and discloses a three-dimensional reconstruction method, a recording medium and a system based on a front scanning sonar image. Background The three-dimensional shape reconstruction of the underwater obstacle is a core technology in the fields of underwater mapping, underwater autonomous navigation and underwater target detection, and has important application value in the scenes of ocean engineering, underwater security protection, hidden detection and the like. The front-sweep and side-sweep sonar is used as main equipment for sensing the underwater environment, can acquire a two-dimensional gray image of an underwater scene by means of the acoustic wave reflection principle, has the advantages of being free from the influence of the turbidity of a water body, long in acting distance, wide in application range and the like, and is an important technical means for realizing the three-dimensional sensing of the underwater unstructured environment. At present, three-dimensional reconstruction technology based on sonar images mostly relies on multi-sensor fusion and multi-viewpoint matching, and the problems of complex system structure, high calculation cost, insufficient instantaneity and the like exist. Meanwhile, the underwater environment is complex and changeable, the factors such as water medium non-uniformity, substrate material difference, sound wave scattering and reverberation interference are easy to cause the phenomena such as artifacts, noise and light and dark area distortion of a sonar image, the traditional two-dimensional image processing method is difficult to effectively distinguish the highlight reflection area, the shadow area and the real obstacle boundary, the three-dimensional morphology resolving precision is insufficient, the contour fitting deviation is large, and the high-precision underwater target reconstruction requirement cannot be met. In practical engineering application, the underwater detection equipment is generally required to have the characteristics of low power consumption, small volume, portable integration, low electromagnetic signal emission and the like so as to adapt to the requirements of long-time underwater operation and concealed detection. Meanwhile, the traditional neural network model has the defects of insufficient characteristic utilization, insufficient fusion of shallow and deep information, limited receptive field range and the like, and the problems of characteristic loss, insufficient segmentation precision, discontinuous reconstruction contour and the like are easy to occur in a complex underwater environment, so that the practical application of the underwater three-dimensional reconstruction technology in hidden detection and low-power-consumption portable equipment is further limited. Therefore, aiming at the problems of poor environmental adaptability, serious artifact and noise interference, insufficient reconstruction precision, large model volume, high power consumption, weak concealment and the like in the existing underwater sonar three-dimensional reconstruction technology, a three-dimensional obstacle shape reconstruction method based on a sonar image, which is light in weight, high in precision, strong in anti-interference capability and suitable for embedded deployment, is needed to meet the actual requirements of high-precision, low-power consumption and hidden three-dimensional perception in a complex underwater environment. Disclosure of Invention Aiming at the problems, the invention provides a three-dimensional reconstruction method based on a front scanning sonar image, which comprises the following steps: s1, acquiring a two-dimensional gray image acquired by ship front scanning sonar; S2, carrying out feature analysis on the two-dimensional gray level image through a semantic segmentation network to obtain a bright-dark feature channel comprising a highlight region and a shadow region, and extracting corresponding boundary features; s3, extracting gray gradient and gray ratio value in an unsupervised manner in a highlight area; s4, nonlinear fitting is carried out by utilizing the gray gradient, the gray ratio and the physical characteristics of sonar reflection, and height contour points of the obstacle along the fore-and-aft direction of the ship body are calculated; s5, fusing boundary features and height contour points of multiple array elements to achieve three-dimensional shape reconstruction of the obstacle. Preferably, the semantic segmentation model in step S2 is a convolutional neural network, based on an original U-net network architecture, in which a hole convolution kernel and residual connection are introduced into a network module, and a parallel structure of separation channels is a