CN112288726A - Method for detecting foreign matters on belt surface of underground belt conveyor - Google Patents
Method for detecting foreign matters on belt surface of underground belt conveyor Download PDFInfo
- Publication number
- CN112288726A CN112288726A CN202011198208.0A CN202011198208A CN112288726A CN 112288726 A CN112288726 A CN 112288726A CN 202011198208 A CN202011198208 A CN 202011198208A CN 112288726 A CN112288726 A CN 112288726A
- Authority
- CN
- China
- Prior art keywords
- belt conveyor
- underground
- image
- foreign matter
- belt surface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000001514 detection method Methods 0.000 claims abstract description 54
- 230000008569 process Effects 0.000 claims abstract description 26
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 230000002708 enhancing effect Effects 0.000 claims abstract description 10
- 238000001914 filtration Methods 0.000 claims description 31
- 238000005286 illumination Methods 0.000 claims description 21
- 230000001965 increasing effect Effects 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000002146 bilateral effect Effects 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000012423 maintenance Methods 0.000 claims description 3
- 230000002787 reinforcement Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 12
- 239000003245 coal Substances 0.000 description 14
- 230000006870 function Effects 0.000 description 8
- 239000002023 wood Substances 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 125000001475 halogen functional group Chemical group 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- -1 anchor rods Substances 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a belt surface foreign matter detection method of an underground belt conveyor, which comprises the following steps of firstly, carrying out image acquisition on belt surface foreign matters of the underground belt conveyor, and constructing a foreign matter data set; secondly, acquiring images of the belt surface of the underground belt conveyor in the operation process, and enhancing and denoising the acquired images; and thirdly, analyzing the foreign matters with the surface by adopting a target detection algorithm based on a convolutional neural network. The method has simple steps, is convenient to realize, can be effectively applied to the belt surface foreign matter detection of the underground belt conveyor, and has high detection accuracy and efficiency, obvious effect and convenient popularization.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a belt surface foreign matter detection method for an underground belt conveyor.
Background
The underground belt conveyor is a key device for coal transportation, but in the coal transportation process, foreign matters such as large coal, anchor rods, wood rods and wood blocks often cause damage to the belt conveyor, so that coal mine units stop production, and huge economic loss is caused. The essence of belt conveyor foreign matter detection is to identify coal and non-coal foreign matters, and in the long-term research and development process, the representative methods include a manual inspection method, a ray method and an image identification method. The manual inspection method is that workers in the well move back and forth to observe whether foreign matters exist on the belt conveyor, and the method requires the workers to observe the belt conveyor by naked eyes for a long time, so that visual fatigue of the workers is easily caused, and potential safety hazards in the well are caused. The detection principle of the ray method is to indirectly identify coal and non-coal foreign matters by utilizing the difference of reflected energy after rays pass through different substances. In practical application, the foreign matter detection system of the underground belt conveyor based on the ray method is expensive and difficult to maintain.
The image recognition method is based on the machine vision, foreign matters threatening belt safety such as underground large coal, anchor rods and wood blocks are analyzed, due to the fact that underground illumination conditions are poor, dust and fog are large, the obtained images are low in illumination and high in noise, underground images need to be enhanced, and underground image enhancement algorithms based on the machine vision are mainly divided into two types, namely underground image enhancement algorithms based on a spatial domain and underground image enhancement algorithms based on a frequency domain. The underground image enhancement algorithm based on the spatial domain processes pixel points in the image according to a certain rule, so that the image is enhanced. The common methods are histogram equalization, logarithmic change and contrast stretching. The algorithm is simple in operation and obvious in effect, but the algorithm does not select information in the image, so that the contrast of an interest area is weakened, and some detailed information is lost. The method comprises the steps of transforming pixels in an image into other spaces through a certain rule by using a frequency domain-based underground image enhancement algorithm, and calculating each pixel point. The common method is an image enhancement algorithm based on wavelet transform, but the method often cannot effectively improve the overall brightness of the image. The single method is difficult to meet the actual requirement, and the enhancement effect is often poor.
Disclosure of Invention
The invention aims to solve the technical problem of providing a belt surface foreign matter detection method of an underground belt conveyor, which has the advantages of simple steps, convenient realization, high detection accuracy and efficiency, obvious effect and convenient popularization, and can be effectively applied to belt surface foreign matter detection of the underground belt conveyor.
In order to solve the technical problems, the invention adopts the technical scheme that: a belt surface foreign matter detection method of an underground belt conveyor comprises the following steps:
firstly, acquiring images of foreign matters on the belt surface of an underground belt conveyor, and constructing a foreign matter data set;
secondly, acquiring images of the belt surface of the underground belt conveyor in the operation process, and enhancing and denoising the acquired images;
and thirdly, analyzing the foreign matters with the surface by adopting a target detection algorithm based on a convolutional neural network.
In the method for detecting the foreign matters on the belt surface of the underground belt conveyor, the specific process of acquiring the images of the foreign matters on the belt surface of the underground belt conveyor and constructing the foreign matter data set in the step one comprises the following steps:
step 101, preliminary selection of a data set;
step 102, expanding a data set;
103, marking a belt surface foreign matter data set of the underground belt conveyor;
and 104, constructing a foreign body training data set and a testing data set of the underground belt conveyor.
In the method for detecting the foreign matter on the belt surface of the underground belt conveyor, the specific process of enhancing and denoising the acquired image in the second step comprises the following steps:
step 201, converting the collected low-illumination image from an RGB color space to an HSV color space;
step 202, enhancing a low-illumination image in an HSV color space by adopting an improved Retinex enhancement method;
step 203, converting the enhanced image from HSV color space to RGB color space;
and step 204, converting the image in the enhanced RGB color space into a gray-scale image, and filtering the gray-scale image by adopting an improved self-adaptive median filtering method.
In the method for detecting foreign matters on the belt surface of the underground belt conveyor, the specific process of the improved Retinex enhancement method in step 202 includes:
a1, estimating the illumination component of the original low-illumination image by adopting bilateral filtering;
a2, performing single-scale Retinex algorithm enhancement on the brightness component, and performing self-adaptive nonlinear stretching on the saturation component to obtain a reflection component with better edge maintenance and an enhanced image for improving the saturation;
a3, obtaining an enhanced brightness component by measuring an index of the reflection component;
and A4, performing contrast correction processing by adopting a global adaptive logarithm enhancement algorithm to obtain a final enhanced image.
In the method for detecting foreign matters on the belt surface of the underground belt conveyor, the specific process of the improved adaptive median filtering method in step 202 includes:
step B1, initializing a filter window SxyMaximum filter window size S of 3max＝13；
Step B2, calculating SxyThe pixel points with the middle pixel value of 0 and 255 are taken as noise points to be removed when S isxyWhen all the pixel points are removed, the size of the filtering window is increased, and when the increased filtering window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is not less than SmaxWhen so, step B3 is performed;
step B3, recalculating SxyMinimum Z of middle pixelminMaximum value ZmaxAnd the median value Zmed；
Step B4, step Zmin＜Zmed＜ZmaxWhen so, step B5 is performed; otherwise, increasing the size of the filter window, and when the increased filter window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is notLess than SmaxTime, output Zxy；
Step B5, step Zmin＜Zxy＜ZmaxTime, output Zxy(ii) a Otherwise, output Zmed。
In the third step, the target detection algorithm based on the convolutional neural network is based on an SSD algorithm frame, the SSD algorithm is optimized by adopting the deep separable convolutional DSC and GIOU, and an extraction layer and default boxes of a feature map in the SSD network are optimized.
The method for detecting the foreign matters on the belt surface of the underground belt conveyor adopts a depth separable convolution DSC to optimize an SSD algorithm, and comprises the following specific processes: and the trunk network VGG16 used for image feature extraction in the SSD algorithm model is simplified by the deep separable convolution DSC, so that the network parameter number is reduced.
The method for detecting the foreign matters on the belt surface of the underground belt conveyor adopts the GIOU to optimize the SSD algorithm, and comprises the following specific steps: replacing the position loss function in the SSD algorithm with GIOU.
The method for detecting the foreign matters on the belt surface of the underground belt conveyor comprises the following specific steps of optimizing an extraction layer and default boxes of a feature map in an SSD network: when feature maps are extracted by using Conv11, Conv13, Conv14, Conv15, Conv16 and Conv17 layers, square default boxes with an aspect ratio of 1' are added.
Compared with the prior art, the invention has the following advantages:
1. the method has simple steps and convenient realization.
2. According to the invention, according to the actual requirement of foreign body detection of the belt conveyor, data acquisition, expansion, marking and other work are carried out, the construction of a foreign body data set of the underground belt conveyor is completed, and the subsequent detection is convenient to carry out.
3. According to the method, the irradiation component is estimated by adopting bilateral filtering, the halo phenomenon generated by the traditional algorithm is effectively avoided, the edge blurring is prevented, the problem of insufficient improvement of the image contrast is solved, the characteristics of a logarithmic function are utilized, the global adaptive logarithm enhancement algorithm is adopted for correction, and the improvement can not affect the high-illumination image. The method reduces the estimation error of the illumination intensity, improves the contrast, and has obvious enhancement effect on the dark area of the low-illumination image, thereby obtaining good visual effect.
4. According to the invention, through an improved self-adaptive median filtering method, noise point detection is carried out before filtering, and the effects of denoising and detail protection can be simultaneously considered.
5. On the basis of an SSD algorithm, the invention introduces a deep separable convolution and a GIOU loss function by optimizing an SSD network structure, thereby improving the detection accuracy and the detection rate.
6. The invention can be effectively applied to the belt surface foreign matter detection of the underground belt conveyor, has high detection accuracy and efficiency, remarkable effect and convenient popularization.
In conclusion, the method has simple steps, is convenient to realize, can be effectively applied to the belt surface foreign matter detection of the underground belt conveyor, and has the advantages of high detection accuracy and efficiency, remarkable effect and convenience in popularization.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a flow chart of a method of improving Retinex enhancement according to the present invention;
FIG. 3 is a diagram showing the effect of the detection test in the detection method of the present invention.
Detailed Description
As shown in fig. 1, the belt surface foreign matter detection method of the underground belt conveyor of the invention comprises the following steps:
firstly, acquiring images of foreign matters on the belt surface of an underground belt conveyor, and constructing a foreign matter data set;
secondly, acquiring images of the belt surface of the underground belt conveyor in the operation process, and enhancing and denoising the acquired images;
and thirdly, analyzing the foreign matters with the surface by adopting a target detection algorithm based on a convolutional neural network.
In this embodiment, in the first step, the specific process of acquiring an image of a foreign object on a belt surface of the underground belt conveyor and constructing a foreign object data set includes:
step 101, preliminary selection of a data set;
during the concrete implementation, in the pit of the colliery, the foreign matter kind is more, can not gather all foreign matter images that probably appear, select often appearing, easily make impaired stock of belt conveyor, the bold coal, the vaulting pole is the representative, carry out data acquisition, in order to satisfy follow-up target detection effect based on convolutional neural network, when designing the data acquisition scheme, need carry out data acquisition according to convolutional neural network to the requirement of data set, including the locating position of foreign matter, the shooting angle of camera, the different kinds of foreign matter, camera exposure intensity, the area has or not the material.
Step 102, expanding a data set;
in specific implementation, the underground coal mine frequently shakes to different degrees, so that the shooting angle of a camera changes, the shaking of a lens is simulated by cutting, rotating, mirroring and other methods when a data set is manufactured, the data set is expanded, and meanwhile, the convolutional neural network is trained by utilizing the expanded data set, so that the robustness of the network can be effectively improved.
103, marking a belt surface foreign matter data set of the underground belt conveyor;
in specific implementation, in order to meet the requirement of position regression during training, the position of the target object in each picture in the data set needs to be labeled.
And 104, constructing a foreign body training data set and a testing data set of the underground belt conveyor.
In specific implementation, the training set and the test set are divided according to the ratio of 8: 2.
In this embodiment, the specific process of enhancing and denoising the acquired image in the second step includes:
step 201, converting the collected low-illumination image from an RGB color space to an HSV color space;
step 202, enhancing a low-illumination image in an HSV color space by adopting an improved Retinex enhancement method;
step 203, converting the enhanced image from HSV color space to RGB color space;
and step 204, converting the image in the enhanced RGB color space into a gray-scale image, and filtering the gray-scale image by adopting an improved self-adaptive median filtering method.
In this embodiment, as shown in fig. 2, the specific process of the improved Retinex enhancement method in step 202 includes:
a1, estimating the illumination component of the original low-illumination image by adopting bilateral filtering;
in specific implementation, the low-illumination image is converted from an RGB color space to an HSV color space, the problems of high dispersity and high correlation of the RGB color space can be solved, bilateral filtering replaces Gaussian filtering as a filtering function, the marginality of the image can be effectively kept, and the halo phenomenon generated by the traditional Retinex algorithm can be avoided; the gaussian similarity weight is related to the magnitude of the pixel value and is the similarity of the pixel value.
In the flat area of the image, the change of the pixel value of the image is small, and the spatial weight plays a main role, namely, Gaussian blurring is performed, while in the edge area of the image, the change range of the pixel value is large, and the similar weight is increased successively, so that the information of the edge is maintained. Therefore, the bilateral filtering can adaptively estimate the place with large edge difference of the low-illumination image, and effectively avoids the halo phenomenon while keeping the image edge information.
A2, performing single-scale Retinex algorithm enhancement on the brightness component, and performing self-adaptive nonlinear stretching on the saturation component to obtain a reflection component with better edge maintenance and an enhanced image for improving the saturation;
a3, obtaining an enhanced brightness component by measuring an index of the reflection component;
and A4, performing contrast correction processing by adopting a global adaptive logarithm enhancement algorithm to obtain a final enhanced image.
When the method is specifically implemented, the global self-adaptive logarithmic enhancement algorithm utilizes logarithmic characteristics and logarithmic mapping relations, logarithmic average brightness is introduced, when the dynamic range of a scene changes, no matter whether an image is too dark or too strong, the average brightness value is always smaller than or equal to the maximum brightness value, the display brightness value can be mapped to 0-1, smooth incremental increase of other display brightness values is ensured, the image cannot be distorted, the logarithmic characteristics are enhanced obviously for low-illumination images, and the logarithmic mapping relations are utilized to prevent the phenomenon of over-enhancement from influencing high-illumination images.
In this embodiment, the specific process of the improved adaptive median filtering method in step 202 includes:
step B1, initializing a filter window SxyMaximum filter window size S of 3max＝13；
Step B2, calculating SxyThe pixel points with the middle pixel value of 0 and 255 are taken as noise points to be removed when S isxyWhen all the pixel points are removed, the size of the filtering window is increased, and when the increased filtering window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is not less than SmaxWhen so, step B3 is performed;
step B3, recalculating SxyMinimum Z of middle pixelminMaximum value ZmaxAnd the median value Zmed；
Step B4, step Zmin＜Zmed＜ZmaxWhen so, step B5 is performed; otherwise, increasing the size of the filter window, and when the increased filter window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is not less than SmaxTime, output Zxy；
Step B5, step Zmin＜Zxy＜ZmaxTime, output Zxy(ii) a Otherwise, output Zmed。
In this embodiment, the target detection algorithm based on the convolutional neural network in step three is based on an SSD algorithm framework, and the SSD algorithm is optimized by using the deep separable convolutional DSC and GIOU, and the extraction layer and default boxes of the feature map in the SSD network are optimized.
In this embodiment, the specific process of optimizing the SSD algorithm by using the depth separable convolution DSC includes: and the trunk network VGG16 used for image feature extraction in the SSD algorithm model is simplified by the deep separable convolution DSC, so that the network parameter number is reduced.
In specific implementation, a backbone network of the SSD algorithm model is VGG16, which is mainly used for feature extraction of images, but the number of network parameters is large, and time consumption is long when image features are extracted.
In this embodiment, the specific process of optimizing the SSD algorithm by using the GIOU includes: replacing the position loss function in the SSD algorithm with GIOU.
In the specific implementation, in the position loss function of the SSD algorithm, the error between the boundary box and the GT image can be determined only by the euclidean distance therebetween, and the size of the overlap region is ignored. In this embodiment, the GIOU is used to replace the position loss function in the original loss function, so as to improve the network detection accuracy.
In this embodiment, the specific process of optimizing the abstraction layer and default boxes of the feature map in the SSD network includes: when feature maps are extracted by using Conv11, Conv13, Conv14, Conv15, Conv16 and Conv17 layers, square default boxes with an aspect ratio of 1' are added.
In specific implementation, in the SSD network, in order to detect target objects with different sizes, feature maps with different sizes are extracted from the layers of Conv4_3, Conv7, Conv8_2, Conv9_2, Conv10_2 and Conv11_2, wherein the large feature map is used for detecting small objects, the small feature map is used for detecting large objects, and default tiles with different aspect ratios are set at each central point of each feature map for detection. In the embodiment, the area ratio of the large coal blocks, the anchor rods, the battens and the wood blocks in the image is large, so that when a feature map is extracted, a small feature map can be extracted for detecting target objects such as the anchor rods and the large coal blocks in a test, and meanwhile, when default boxes are generated, square default boxes with the length-width ratio of 1' are added, and the network detection accuracy is improved.
In order to verify the effect of the method, the target detection algorithm is compared with other deep learning detection algorithms in tests on the detection accuracy and the detection rate. During training, the value of batch _ size is 32, the value of lr is 0.001, the value of epoch is 30000, the types of the large coal, the anchor rod, the wood rod and the wood block are respectively set to be numbers of '1', '2', '3' and '4', and after training, the detection accuracy and the detection rate of the foreign matters with the surface are compared as shown in table 1.
TABLE 1 comparison of the results of the object detection algorithm of the present invention with other deep learning detection algorithms
As can be seen from Table 1, compared with other deep learning detection algorithms, the target detection algorithm has advantages in detection accuracy and detection rate, and compared with the original SSD algorithm, the target detection algorithm has higher detection rate, and the detection efficiency is greatly improved as the original 32 frames per second is processed to 41 frames per second; meanwhile, the detection accuracy is improved from 87.1% to 90.2%, and the foreign matter detection capability on the belt surface of the belt conveyor is greatly improved.
FIG. 3 shows the effect of the detection test using the detection method of the present invention.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and all simple modifications, changes and equivalent structural changes made to the above embodiment according to the technical spirit of the present invention still fall within the protection scope of the technical solution of the present invention.
Claims (9)
1. A belt surface foreign matter detection method of an underground belt conveyor is characterized by comprising the following steps:
firstly, acquiring images of foreign matters on the belt surface of an underground belt conveyor, and constructing a foreign matter data set;
secondly, acquiring images of the belt surface of the underground belt conveyor in the operation process, and enhancing and denoising the acquired images;
and thirdly, analyzing the foreign matters with the surface by adopting a target detection algorithm based on a convolutional neural network.
2. The method for detecting the foreign matter on the belt surface of the underground belt conveyor according to claim 1, wherein the specific process of acquiring the image of the foreign matter on the belt surface of the underground belt conveyor and constructing the foreign matter data set in the step one comprises the following steps:
step 101, preliminary selection of a data set;
step 102, expanding a data set;
103, marking a belt surface foreign matter data set of the underground belt conveyor;
and 104, constructing a foreign body training data set and a testing data set of the underground belt conveyor.
3. The method for detecting the foreign matter on the belt surface of the underground belt conveyor as claimed in claim 1, wherein the specific process of enhancing and denoising the acquired image in the second step comprises:
step 201, converting the collected low-illumination image from an RGB color space to an HSV color space;
step 202, enhancing a low-illumination image in an HSV color space by adopting an improved Retinex enhancement method;
step 203, converting the enhanced image from HSV color space to RGB color space;
and step 204, converting the image in the enhanced RGB color space into a gray-scale image, and filtering the gray-scale image by adopting an improved self-adaptive median filtering method.
4. The method for detecting the foreign matters on the belt surface of the underground belt conveyor as claimed in claim 3, wherein the specific process of the improved Retinex reinforcement method in the step 202 comprises the following steps:
a1, estimating the illumination component of the original low-illumination image by adopting bilateral filtering;
a2, performing single-scale Retinex algorithm enhancement on the brightness component, and performing self-adaptive nonlinear stretching on the saturation component to obtain a reflection component with better edge maintenance and an enhanced image for improving the saturation;
a3, obtaining an enhanced brightness component by measuring an index of the reflection component;
and A4, performing contrast correction processing by adopting a global adaptive logarithm enhancement algorithm to obtain a final enhanced image.
5. The method for detecting foreign matter on the belt surface of a downhole belt conveyor according to claim 3, wherein the specific process of the improved adaptive median filtering method in step 202 comprises:
step B1, initializing a filter window SxyMaximum filter window size S of 3max＝13；
Step B2, calculating sxyThe pixel points with the middle pixel value of 0 and 255 are taken as noise points to be removed when S isxyWhen all the pixel points are removed, the size of the filtering window is increased, and when the increased filtering window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is not less than SmaxWhen so, step B3 is performed;
step B3, recalculating SxyMinimum Z of middle pixelminMaximum value ZmaxAnd the median value Zmed；
Step B4, step Zmin＜Zmed＜ZmaxWhen so, step B5 is performed; otherwise, increasing the size of the filter window, and when the increased filter window is smaller than SmaxWhen so, step B2 is performed; when the increased filtering window is not less than SmaxTime, output Zxy；
Step B5, step Zmin＜Zxy＜ZmaxTime, output Zxy(ii) a Otherwise, output Zmed。
6. The method for detecting the foreign matter on the belt surface of the underground belt conveyor according to claim 1, wherein the target detection algorithm based on the convolutional neural network in the third step is based on an SSD algorithm framework, the SSD algorithm is optimized by adopting a deep separable convolutional DSC and GIOU, and an extraction layer and default boxes of a feature map in the SSD network are optimized.
7. The method for detecting foreign matter on the belt surface of a downhole belt conveyor according to claim 6, wherein the specific process of optimizing the SSD algorithm by using the depth separable convolution DSC comprises the following steps: and the trunk network VGG16 used for image feature extraction in the SSD algorithm model is simplified by the deep separable convolution DSC, so that the network parameter number is reduced.
8. The method for detecting foreign matter on the belt surface of a downhole belt conveyor according to claim 6, wherein the specific process of optimizing the SSD algorithm by using GIOU comprises the following steps: replacing the position loss function in the SSD algorithm with GIOU.
9. The method for detecting the foreign matters on the belt surface of the underground belt conveyor as claimed in claim 6, wherein the specific process of optimizing the extraction layer and the default boxes of the feature map in the SSD network comprises the following steps: when feature maps are extracted by using Conv11, Conv13, Conv14, Conv15, Conv16 and Conv17 layers, square default boxes with an aspect ratio of 1' are added.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011198208.0A CN112288726B (en) | 2020-10-30 | 2020-10-30 | Method for detecting foreign matters on belt surface of underground belt conveyor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011198208.0A CN112288726B (en) | 2020-10-30 | 2020-10-30 | Method for detecting foreign matters on belt surface of underground belt conveyor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112288726A true CN112288726A (en) | 2021-01-29 |
| CN112288726B CN112288726B (en) | 2023-12-29 |
Family
ID=74353522
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011198208.0A Active CN112288726B (en) | 2020-10-30 | 2020-10-30 | Method for detecting foreign matters on belt surface of underground belt conveyor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112288726B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022178833A1 (en) * | 2021-02-26 | 2022-09-01 | 京东方科技集团股份有限公司 | Target detection network training method, target detection method, and apparatus |
| CN117292212A (en) * | 2023-11-27 | 2023-12-26 | 太原理工大学 | Foreign matter identification method of scraper conveyor |
| US12002254B2 (en) | 2021-02-26 | 2024-06-04 | Boe Technology Group Co., Ltd. | Method and apparatus of training object detection network and object detection method and apparatus |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080101719A1 (en) * | 2006-10-30 | 2008-05-01 | Samsung Electronics Co., Ltd. | Image enhancement method and system |
| RU2622095C1 (en) * | 2016-02-05 | 2017-06-09 | Акционерное общество "Государственный Рязанский приборный завод" | Method of improving digital colour images |
| CN110675343A (en) * | 2019-09-24 | 2020-01-10 | 西安科技大学 | Image enhancement method for underground coal conveying image |
| CN111429370A (en) * | 2020-03-23 | 2020-07-17 | 煤炭科学技术研究院有限公司 | Method and system for enhancing images in coal mine and computer storage medium |
| CN111612758A (en) * | 2020-05-19 | 2020-09-01 | 南京航空航天大学 | Tunnel surface defect detection method based on convolutional neural network |
| CN111833366A (en) * | 2020-06-03 | 2020-10-27 | 佛山科学技术学院 | Edge detection method based on Canny algorithm |
-
2020
- 2020-10-30 CN CN202011198208.0A patent/CN112288726B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080101719A1 (en) * | 2006-10-30 | 2008-05-01 | Samsung Electronics Co., Ltd. | Image enhancement method and system |
| RU2622095C1 (en) * | 2016-02-05 | 2017-06-09 | Акционерное общество "Государственный Рязанский приборный завод" | Method of improving digital colour images |
| CN110675343A (en) * | 2019-09-24 | 2020-01-10 | 西安科技大学 | Image enhancement method for underground coal conveying image |
| CN111429370A (en) * | 2020-03-23 | 2020-07-17 | 煤炭科学技术研究院有限公司 | Method and system for enhancing images in coal mine and computer storage medium |
| CN111612758A (en) * | 2020-05-19 | 2020-09-01 | 南京航空航天大学 | Tunnel surface defect detection method based on convolutional neural network |
| CN111833366A (en) * | 2020-06-03 | 2020-10-27 | 佛山科学技术学院 | Edge detection method based on Canny algorithm |
Non-Patent Citations (6)
| Title |
|---|
| YUANBIN WANG等: "Video detection of foreign objects on the surface of belt conveyor underground coal mine based on improved SSD", 《JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING》 * |
| 付燕等: "一种煤矿井下视频图像增强算法", 《工矿自动化》 * |
| 刘松等: "基于单尺度Retinex雾天降质图像增强算法", 《微电子学与计算机》 * |
| 吴守鹏等: "基于改进FPN的输送带异物识别方法", 《煤矿安全》 * |
| 董静薇等: "基于同态滤波及多尺度Retinex的低照度图像增强算法", 《科学技术与工程》 * |
| 钱伟新等: "一种改进的自适应中值滤波算法", 《光学与光电技术》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022178833A1 (en) * | 2021-02-26 | 2022-09-01 | 京东方科技集团股份有限公司 | Target detection network training method, target detection method, and apparatus |
| US12002254B2 (en) | 2021-02-26 | 2024-06-04 | Boe Technology Group Co., Ltd. | Method and apparatus of training object detection network and object detection method and apparatus |
| CN117292212A (en) * | 2023-11-27 | 2023-12-26 | 太原理工大学 | Foreign matter identification method of scraper conveyor |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112288726B (en) | 2023-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110349126B (en) | Convolutional neural network-based marked steel plate surface defect detection method | |
| CN108921799B (en) | Remote sensing image thin cloud removing method based on multi-scale collaborative learning convolutional neural network | |
| CN110490914B (en) | Image fusion method based on brightness self-adaption and significance detection | |
| CN108876743B (en) | Image rapid defogging method, system, terminal and storage medium | |
| CN107230202B (en) | Automatic identification method and system for road surface disease image | |
| CN112819772B (en) | High-precision rapid pattern detection and recognition method | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN111062293B (en) | Unmanned aerial vehicle forest flame identification method based on deep learning | |
| CN109447917B (en) | Remote sensing image haze eliminating method based on content, characteristics and multi-scale model | |
| Saini et al. | Object detection in underwater image by detecting edges using adaptive thresholding | |
| CN114118144A (en) | Anti-interference accurate aerial remote sensing image shadow detection method | |
| CN110889837A (en) | Cloth flaw detection method with flaw classification function | |
| CN111709888B (en) | Aerial image defogging method based on improved generation countermeasure network | |
| CN108664939A (en) | A kind of remote sensing images aircraft recognition method based on HOG features and deep learning | |
| CN114926407A (en) | Steel surface defect detection system based on deep learning | |
| CN111754538B (en) | Threshold segmentation method for USB surface defect detection | |
| CN110969620A (en) | Method and device for detecting magnetic shoe ripple defects | |
| CN112288726B (en) | Method for detecting foreign matters on belt surface of underground belt conveyor | |
| CN113505702A (en) | Pavement disease identification method and system based on double neural network optimization | |
| CN109726660A (en) | A kind of remote sensing images ship identification method | |
| CN117274113B (en) | Broken silicon wafer cleaning effect visual detection method based on image enhancement | |
| CN113421210A (en) | Surface point cloud reconstruction method based on binocular stereo vision | |
| CN117333359A (en) | Mountain-water painting image super-resolution reconstruction method based on separable convolution network | |
| CN104102911A (en) | Image processing for AOI (automated optical inspection)-based bullet appearance defect detection system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20231127 Address after: Room 1106, Building 2, Rongmin Central International, No. 99 Weiyang Road, Lianhu District, Xi'an City, Shaanxi Province, 710000 Applicant after: Xi'an zhicaiquan Technology Transfer Center Co.,Ltd. Address before: 710054 No. 58, middle section, Yanta Road, Shaanxi, Xi'an Applicant before: XI'AN University OF SCIENCE AND TECHNOLOGY |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |