CN-117975222-B - Feature enhancement-based image panorama segmentation method, device, equipment and storage medium

CN117975222BCN 117975222 BCN117975222 BCN 117975222BCN-117975222-B

Abstract

The invention discloses a feature enhancement-based image panorama segmentation method, device and equipment and a storage medium, and belongs to the technical field of computer vision identification. The method comprises the steps of preprocessing an original image to obtain a processed original image, inputting the processed original image into an image processing network to extract image features to obtain a multi-level feature image set, carrying out image feature enhancement based on the multi-level feature image set to obtain an enhanced feature image set, carrying out target feature redistribution on the enhanced feature image set to obtain a target feature image set, and carrying out panoramic segmentation on the target feature image set to obtain a panoramic segmented image. According to the invention, the processed original image is input into an image processing network for feature extraction, feature enhancement and target feature redistribution, so that a panoramic segmentation image corresponding to the original image is generated, and the understanding of activities and events in an image scene and the detection and identification accuracy of a target object are improved.

Inventors

YUAN XIAOCHEN
JI SHUCHENG
LIN CANTANG
YAN ZHAOJI

Assignees

澳门理工大学

Dates

Publication Date: 20260512
Application Date: 20240306

Claims (9)

1. A feature enhancement-based image panorama segmentation method, comprising: Preprocessing an original image to obtain a processed original image; Inputting the processed original image into an image processing network for image feature extraction to obtain a multi-level feature image set; image feature enhancement is carried out based on the multi-level feature image set, and an enhanced feature image set is obtained; performing target feature redistribution on the enhanced feature image set to obtain a target feature image set, wherein the target feature image set specifically comprises: Carrying out feature image combination on the enhanced feature image set to obtain an overall-stage feature image; Performing target feature weight calculation based on the enhanced feature image set to obtain stage attention weight data; Performing integration processing based on the integral stage feature image and the stage attention weight data to obtain the target feature image set; And carrying out panoramic segmentation processing on the target characteristic image set to obtain a panoramic segmented image.
2. The feature enhancement-based image panorama segmentation method according to claim 1, wherein the image feature enhancement is performed based on the multi-level feature image set to obtain an enhanced feature image set, and the method specifically comprises: feature fusion is respectively carried out on the feature images of different layers in the multi-layer feature image set according to the scale size, and a fused feature image set is obtained; and carrying out image feature enhancement on the fused feature image set to obtain the enhanced feature image set.
3. The feature enhancement-based image panorama segmentation method according to claim 2, wherein the image feature enhancement is performed based on the multi-level feature image set, and further comprising, after obtaining the enhanced feature image set: and generating a convolution kernel based on the enhanced characteristic image set to obtain an image panoramic segmentation convolution kernel.
4. The feature enhancement-based image panorama segmentation method according to claim 3, wherein the panorama segmentation process is performed on the target feature image set to obtain a panorama segmented image, and the method specifically comprises: performing integrated coding processing on the target feature image set to obtain a coded target feature image; decoding the coded target characteristic image to obtain a panoramic segmentation result; and carrying out image post-processing according to the panoramic segmentation result to obtain the panoramic segmentation image.
5. The feature enhancement-based image panorama segmentation method according to claim 4, wherein the performing an integration encoding process on the target feature image set to obtain an encoded target feature image comprises: extracting coding features of the target feature image set to obtain a space coding feature image; attention weight distribution is carried out according to the space coding feature images, so that unified coding feature images are obtained; And checking the unified coding feature image based on the image panoramic segmentation convolution to carry out segmentation prediction so as to obtain a coded target feature image.
6. The feature enhancement-based image panorama segmentation method according to claim 5, wherein decoding the encoded target feature image to obtain a panorama segmentation result comprises: Performing up-sampling processing on the coded target feature image to obtain a sampled target feature image; And carrying out feature fusion processing on the sampled target feature image to obtain the panoramic segmentation result.
7. An image panorama dividing device based on characteristic enhancement is characterized in that, the feature enhancement-based image panorama segmentation apparatus comprises: The preprocessing module is used for preprocessing the original image to obtain a processed original image; the feature extraction module is used for inputting the processed original image into an image processing network to extract image features so as to obtain a multi-level feature image set; the characteristic enhancement module is used for enhancing the image characteristic based on the multi-level characteristic image set to obtain an enhanced characteristic image set; The feature redistribution module is used for performing target feature redistribution on the enhanced feature image set to obtain a target feature image set, and specifically comprises the following steps: Carrying out feature image combination on the enhanced feature image set to obtain an overall-stage feature image; Performing target feature weight calculation based on the enhanced feature image set to obtain stage attention weight data; Performing integration processing based on the integral stage feature image and the stage attention weight data to obtain the target feature image set; And the panoramic segmentation module is used for carrying out panoramic segmentation processing on the target characteristic image set to obtain a panoramic segmented image.
8. A feature-enhancement-based image panorama segmentation apparatus, comprising a memory, a processor, and a feature-enhancement-based image panorama segmentation program stored on the memory and executable on the processor, the feature-enhancement-based image panorama segmentation program configured to implement the feature-enhancement-based image panorama segmentation method according to any one of claims 1-6.
9. A computer readable storage medium storing a computer program, characterized in that the computer program is capable of realizing the steps in the feature-enhanced image panorama segmentation method according to any one of claims 1 to 6 when executed by a processor.

Description

Feature enhancement-based image panorama segmentation method, device, equipment and storage medium Technical Field The present invention relates to the field of computer vision recognition technologies, and in particular, to a feature enhancement-based image panorama segmentation method, device, apparatus, and storage medium. Background With the continuous deep research of computer vision and deep learning methods, technologies such as image classification, semantic segmentation, instance segmentation and the like based on deep learning have all obtained great progress. Semantic segmentation assigns a semantic class label to each pixel in an image, but does not distinguish between different object instances of the same semantic class in the image. Example segmentation performs pixel-level segmentation on object examples in an image, but does not involve various non-countable objects that have no explicit shape. The panoramic segmentation task is the unification of semantic segmentation and instance segmentation tasks, and is very important for tasks such as automatic driving and intelligent robots which rely on visual perception of image scenes. The prior art may not effectively apply the attention mechanism and the coordinate convolution, thus having shortcomings in redistributing a large amount of channel information and providing accurate spatial information, resulting in reduced performance when dealing with small objects or scenes with rich detail. In view of the foregoing, there is a need for an RGB image panorama segmentation prediction optimization method implemented by using multiple feature enhancement techniques, which is used in the fields of automatic driving and video monitoring, so as to improve understanding of activities and events in a scene, and better implement detection and identification of a target object. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The invention mainly aims to provide an image panorama segmentation method, device, equipment and storage medium based on feature enhancement, and aims to solve the technical problem that in the prior art, the accuracy is not high when image panorama segmentation is carried out. To achieve the above object, the present invention provides a feature enhancement-based image panorama segmentation method, comprising the steps of: Preprocessing an original image to obtain a processed original image; Inputting the processed original image into an image processing network for image feature extraction to obtain a multi-level feature image set; image feature enhancement is carried out based on the multi-level feature image set, and an enhanced feature image set is obtained; Performing target feature redistribution on the enhanced feature image set to obtain a target feature image set; And carrying out panoramic segmentation processing on the target characteristic image set to obtain a panoramic segmented image. Optionally, image feature enhancement is performed based on the multi-level feature image set, so as to obtain an enhanced feature image set, which specifically includes: feature fusion is respectively carried out on the feature images of different layers in the multi-layer feature image set according to the scale size, and a fused feature image set is obtained; and carrying out image feature enhancement on the fused feature image set to obtain the enhanced feature image set. Optionally, image feature enhancement is performed based on the multi-level feature image set, and after the enhanced feature image set is obtained, the method further includes: and generating a convolution kernel based on the enhanced characteristic image set to obtain an image panoramic segmentation convolution kernel. Optionally, performing target feature redistribution on the enhanced feature image set to obtain a target feature image set, which specifically includes: Carrying out feature image combination on the enhanced feature image set to obtain an overall-stage feature image; Performing target feature weight calculation based on the enhanced feature image set to obtain stage attention weight data; And carrying out integration processing based on the integral stage characteristic image and the stage attention weight data to obtain the target characteristic image set. Optionally, performing panoramic segmentation processing on the target feature image set to obtain a panoramic segmented image, which specifically includes: performing integrated coding processing on the target feature image set to obtain a coded target feature image; decoding the coded target characteristic image to obtain a panoramic segmentation result; and carrying out image post-processing according to the panoramic segmentation result to obtain the panoramic segmentation image. Optionally, performing an inte