CN-122023972-A - Underground anomaly intelligent recognition method and system based on image enhancement

Abstract

The invention discloses an image enhancement-based underground anomaly intelligent recognition method and system, and relates to the technical field of image enhancement. The method comprises the steps of obtaining an underground image, extracting image features, calculating dust shielding coefficients based on the image features and air sensing data, conducting self-adaptive image enhancement to obtain an enhanced image, conducting primary anomaly identification based on the enhanced image and a regional historical image set, conducting image segmentation and secondary enhancement based on a primary identification result and a historical image set to obtain an enhanced segmented image set and risk factors, and finally conducting specific identification and risk correction by combining the risk factors to output an underground anomaly intelligent identification result. According to the method, accurate image enhancement is realized through dynamic perception of environmental changes, the historical data is comprehensively utilized to conduct anomaly positioning and risk quantification, and the accuracy, reliability and early warning timeliness of anomaly identification in a complex underground environment are improved.

Inventors

• XU ZHIOU
• WANG HU

Assignees

• 中国矿业大学

Dates

Publication Date

20260512

Application Date

20260211

Claims (10)

1. 1. The image enhancement-based underground anomaly intelligent identification method is characterized by comprising the following steps of: Acquiring an underground image and extracting image characteristics of the underground image; acquiring a dust shielding coefficient based on the image characteristics and the air sensing data, and performing image enhancement on the underground image to acquire an enhanced image; based on the enhanced image, combining a regional historical image set to perform preliminary abnormal recognition, and acquiring a preliminary recognition result; And carrying out image segmentation based on the primary identification result and combining with the regional historical image set, carrying out secondary enhancement, obtaining an enhanced segmented image set and risk factors, carrying out specific identification, and obtaining an underground abnormal intelligent identification result.
2. 2. The image-enhancement-based intelligent identification method for downhole anomalies according to claim 1, wherein acquiring a downhole image and extracting image features of the downhole image comprises: Acquiring an underground image; Extracting image features of the downhole image, wherein the image features comprise color features, texture features and edge gradient features.
3. 3. The image enhancement-based intelligent identification method for downhole anomalies according to claim 1, wherein acquiring a dust shielding coefficient based on the image features and air sensing data, performing image enhancement on the downhole image, and acquiring an enhanced image, comprises: Inputting the image features into an image feature identifier to obtain an image shielding coefficient; acquiring air sensing data of the area where the underground image is positioned, and calculating and acquiring a dust shielding coefficient by combining the image shielding coefficient; and matching an image enhancement strategy based on the dust shielding coefficient, and carrying out image enhancement on the underground image by adopting the image enhancement strategy to obtain an enhanced image.
4. 4. The method for intelligently identifying the downhole anomalies based on the image enhancement as set forth in claim 3, wherein the acquisition of the image feature identifier includes: acquiring a historical image set, and acquiring image characteristics of each historical image in the historical image set as a sample input set; Performing image shielding degree labeling on the historical image to obtain a sample image shielding coefficient set, wherein the image shielding degree labeling is obtained based on the degradation degree of the image characteristics; based on machine learning, an image feature identifier is constructed, and the image feature identifier is trained by adopting the sample input set and the sample image shielding coefficient set until convergence.
5. 5. The image-enhancement-based underground anomaly intelligent recognition method according to claim 1, wherein the step of performing preliminary anomaly recognition based on the enhanced image in combination with the regional history image set to obtain a preliminary recognition result comprises the following steps: acquiring a regional historical image set and acquiring a normal position range of equipment; Combining the normal position range of the equipment, carrying out position abnormality recognition on the enhanced image, and obtaining a position abnormality recognition result; based on the regional historical image set and the image features of the enhanced image, carrying out feature statistics deviation recognition, obtaining a feature abnormality recognition result and obtaining a feature abnormality recognition result; and combining the position abnormality recognition result and the characteristic abnormality recognition result to obtain the preliminary recognition result.
6. 6. The image-enhancement-based intelligent identification method for downhole anomalies according to claim 1, wherein image segmentation is performed based on the preliminary identification result in combination with a region history image set, comprising: Acquiring a preliminary abnormal position based on the preliminary identification result; image segmentation is carried out based on the preliminary abnormal position, segmented images are obtained, and the segmented images are added into the segmented image set; And based on the high-frequency abnormal positions in the regional history image set, performing high-frequency abnormal position segmentation on the enhanced image, acquiring a segmented image, and adding the segmented image into the segmented image set.
7. 7. The method for intelligently identifying an anomaly downhole based on image enhancement of claim 6, wherein obtaining the high frequency anomaly location comprises: acquiring abnormal conditions in the regional historical image set; Acquiring an abnormal frequency based on the occurrence frequency of the abnormal condition in the historical image set; Combining the historical dust shielding coefficient of the historical image set to obtain an abnormal frequency threshold value, and screening the abnormal condition to obtain a high-frequency abnormality; and extracting an abnormal position of the high-frequency abnormality as the high-frequency abnormal position.
8. 8. The method for intelligently identifying the underground abnormality based on the image enhancement according to claim 1, wherein the steps of performing secondary enhancement, obtaining an enhanced segmented image set and risk factors, performing specific identification, and obtaining an intelligent identification result of the underground abnormality comprise the following steps: combining the primary recognition result, carrying out secondary enhancement on the images in the segmented image set, and obtaining an enhanced segmented image set; Acquiring a risk factor based on the image characteristics of the enhanced segmented image set and a preliminary identification result, wherein the risk factor comprises an anomaly type and an anomaly risk, the anomaly type is acquired based on the preliminary identification result, and the anomaly risk is acquired based on the anomaly frequency of the nearest anomaly position of the enhanced segmented image; and combining the risk factors to complete the branch identification matching and perform specific identification, and obtaining the underground abnormal intelligent identification result.
9. 9. The image-enhancement-based intelligent identification method for downhole anomalies according to claim 1, wherein the step of completing identification branch matching and performing specific identification in combination with the risk factors to obtain the intelligent identification result for downhole anomalies comprises the following steps: Acquiring a downhole anomaly identification model, wherein the downhole anomaly identification model comprises a plurality of anomaly identification branches, and each anomaly identification branch is trained and acquired based on a historical anomaly image set and a historical anomaly annotation set of an anomaly type; Inputting the enhanced segmentation image into the underground anomaly recognition model, carrying out recognition branch matching based on the risk factors, and carrying out specific recognition to obtain a specific recognition result, wherein the specific recognition result comprises an underground anomaly type and an underground anomaly risk; And carrying out risk correction on the specific identification result based on the risk factor, and integrating the underground abnormal type and corrected underground abnormal risk into an underground abnormal intelligent identification result.
10. 10. An image-enhancement-based intelligent downhole anomaly identification system for performing the image-enhancement-based intelligent downhole anomaly identification method of any one of claims 1-9, comprising: the image acquisition module is used for acquiring an underground image and extracting image characteristics of the underground image; The image enhancement module is used for acquiring a dust shielding coefficient based on the image characteristics and the air sensing data, and performing image enhancement on the underground image to acquire an enhanced image; the preliminary anomaly identification module is used for carrying out preliminary anomaly identification based on the enhanced image and combining a regional historical image set to obtain a preliminary identification result; And the specificity recognition module is used for carrying out image segmentation based on the primary recognition result and combining with the regional historical image set, carrying out secondary enhancement, obtaining an enhanced segmented image set and risk factors, carrying out specificity recognition, and obtaining an underground abnormal intelligent recognition result.

Description

Underground anomaly intelligent recognition method and system based on image enhancement Technical Field The invention relates to the technical field of image enhancement, in particular to an intelligent identification method and system for underground anomalies based on image enhancement. Background The underground environment has the inherent challenges of insufficient illumination, diffuse dust and the like, and directly acquired images are often low in contrast, fuzzy in detail and serious in noise interference. The existing underground image anomaly recognition scheme focuses on directly applying a general image enhancement algorithm or a target detection model, generally performs overall unified processing on the image, and fails to fully consider the differential influence of underground dust concentration dynamic change on the degradation degree of the image, and has limited enhancement effect and insufficient adaptability. Meanwhile, the existing method only relies on single-frame image information for analysis, and lacks fusion utilization of historical states and abnormal modes of a monitoring area, so that the sensing capability of abnormal features such as fine change of textures, small targets or position deviation and the like is weak, the recognition result has high false alarm rate and false alarm rate, and the actual requirement of high-reliability safety early warning is difficult to meet. Disclosure of Invention Aiming at the technical problems that the downhole image enhancement method in the prior art is poor in adaptability, history information is difficult to fully utilize and the recognition accuracy of complex anomalies is insufficient, the invention provides the downhole anomaly intelligent recognition method and system based on image enhancement. The technical scheme for solving the technical problems is as follows: In a first aspect, the present invention provides an image enhancement-based method for intelligently identifying downhole anomalies, comprising: Acquiring an underground image and extracting image characteristics of the underground image; acquiring a dust shielding coefficient based on the image characteristics and the air sensing data, and performing image enhancement on the underground image to acquire an enhanced image; based on the enhanced image, combining a regional historical image set to perform preliminary abnormal recognition, and acquiring a preliminary recognition result; And carrying out image segmentation based on the primary identification result and combining with the regional historical image set, carrying out secondary enhancement, obtaining an enhanced segmented image set and risk factors, carrying out specific identification, and obtaining an underground abnormal intelligent identification result. In a second aspect, the present invention provides an image enhancement-based downhole anomaly intelligent recognition system, comprising: the image acquisition module is used for acquiring an underground image and extracting image characteristics of the underground image; The image enhancement module is used for acquiring a dust shielding coefficient based on the image characteristics and the air sensing data, and performing image enhancement on the underground image to acquire an enhanced image; the preliminary anomaly identification module is used for carrying out preliminary anomaly identification based on the enhanced image and combining a regional historical image set to obtain a preliminary identification result; And the specificity recognition module is used for carrying out image segmentation based on the primary recognition result and combining with the regional historical image set, carrying out secondary enhancement, obtaining an enhanced segmented image set and risk factors, carrying out specificity recognition, and obtaining an underground abnormal intelligent recognition result. The beneficial effects of the invention are as follows: Compared with the prior art, the method and the device for automatically and dynamically calculating the dust shielding coefficient by fusing the image characteristics and the real-time air sensing data and performing self-adaptive image enhancement effectively solve the problem of unstable image quality caused by underground dust concentration change. Secondly, on the basis of first enhancement, a regional historical image set is introduced as a reference standard, and preliminary abnormal positioning is realized through comparison and analysis, so that environmental background interference is reduced. And performing refined image segmentation based on the preliminary result and the historical high-frequency abnormal information, and performing secondary targeted enhancement on the segmented region, so that the visual expressive force and the characteristic separability of the key abnormal region are improved. Finally, combining the calculated risk factors, guiding a special recognition branch model to perfor