CN114299290B - Bare soil identification method, device, equipment and computer readable storage medium - Google Patents
Bare soil identification method, device, equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN114299290B CN114299290B CN202111601745.XA CN202111601745A CN114299290B CN 114299290 B CN114299290 B CN 114299290B CN 202111601745 A CN202111601745 A CN 202111601745A CN 114299290 B CN114299290 B CN 114299290B
- Authority
- CN
- China
- Prior art keywords
- bare soil
- image
- pixel
- bare
- deleting
- 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.)
- Active
Links
Images
Abstract
The invention discloses a bare soil identification method, a bare soil identification device, bare soil identification equipment and a computer readable storage medium. The method comprises inputting an image to be detected; performing pixel level prediction on the image through a deep learning neural network pre-training model, and performing image segmentation; setting a confidence threshold value to filter the image; extracting pixels identified as bare soil as an image mask, and performing feature extraction on the image by using the image mask; carrying out secondary classification on the bare soil and the cement through colors, and deleting pixels which are judged to be non-bare soil in the image; reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity; and dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks. The invention is applied to the technical field of image processing.
Description
Technical Field
The present invention relates to the field of image processing, and in particular, to a method, an apparatus, a device, and a computer-readable storage medium for bare soil identification.
Background
Today, semantic segmentation (applied to still 2D images, video and even 3D data, volumetric data) is one of the key issues of computer vision. In a macroscopic sense, semantic segmentation is a high-level task that paves roads for scene understanding. As a core problem of computer vision, scene understanding is increasingly important, because more and more application scenes in reality need to infer relevant knowledge or semantics (i.e., concrete to abstract processes) from images. These applications include autopilot, human-computer interaction, computational photography, image search engines, augmented reality, and the like. These problems have been solved using various conventional computer vision and machine learning techniques. Although these methods are popular, the deep learning revolution has led to a change in the relevant fields, and therefore many computer vision problems including semantic segmentation are beginning to be solved by using a deep architecture, usually adopting a convolutional neural network CNN, which is far more accurate and efficient than the conventional methods. However, deep learning is far less mature than the inherent computer vision and machine learning branches. Therefore, there is no uniform work and no overview of the current optimal approach. The rapid development in this field makes it difficult to enlighten education for beginners, and it is very time-consuming to keep pace with the development as a large amount of work is put forward one after another. It is then very difficult to follow semantic segmentation related efforts, to reasonably interpret their arguments, to filter low-level efforts, and to validate related experimental results, etc. More specifically, the goal of semantic image segmentation is to label each pixel of the image with the corresponding represented class. This task is often referred to as dense prediction because we are predicting every pixel in the image. And the traditional machine learning has the characteristics of high operation speed and simple design. In a simple classification problem, a large number of features are not needed, and only a few main features are selected to obtain better classification performance.
Bare soil is often generated in the urban construction process, and if the bare soil cannot be treated in time, dust, muddy water and the like are easily generated, so that the urban environment is greatly polluted, and the public health is invaded, therefore, supervision workers are required to cover the bare soil with blue cloth in the actual construction site operation. Under the scenes related to urban construction, only the types of objects like bare soil need to be identified, but in the prior art, the bare soil is identified by retraining the neural network model, but the retraining of the neural network model needs to reprocess and label data, so that the development process is complex, the false detection rate is high, most of false detection parts are cement, and therefore, a bare soil identification method, a device, equipment and a computer readable storage medium with short development time, high classification speed and high accuracy need to be developed at present.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defects of the prior art, and the first purpose is to provide a bare soil identification method, a device, equipment and a computer readable storage medium with short development time, high classification speed and high accuracy.
The invention aims to provide the bare soil recognition device which is short in development time, high in classification speed and high in accuracy.
The third purpose of the invention is to provide bare soil identification equipment with short development time, high classification speed and high accuracy.
A fourth object of the present invention is to provide a computer-readable storage medium with short development time, fast classification speed and high accuracy.
The invention provides a soil identification method, which comprises the following steps:
inputting an image to be detected;
performing pixel level prediction on the image through a deep learning neural network pre-training model, and performing image segmentation;
setting a confidence threshold value to filter the image;
extracting pixels identified as bare soil as an image mask, and performing feature extraction on the image by using the image mask;
carrying out secondary classification on bare soil and cement through colors, and deleting pixels which are judged to be non-bare soil in the image;
reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity;
and according to the number of the complete pixel blocks, dividing the predicted image into two types of bare soil and bare soil.
Further, the step of setting a confidence threshold to filter the image includes:
and (4) taking the confidence degree threshold value as 0.9, judging the soil to be bare soil if the pixel confidence degree value is greater than 0.9, and otherwise, predicting the result type according to the confidence degree forward delay.
Further, the step of classifying the bare soil and the cement twice by colors and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
setting a chromaticity threshold value to classify bare soil and cement twice, and deleting pixels which are judged to be non-bare soil in the image;
and setting a color ratio threshold value to carry out secondary classification on the bare soil and the cement, and deleting pixels which are judged to be non-bare soil in the image.
Further, the step of setting the chromaticity threshold value to classify bare soil and cement twice and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
and taking the chroma threshold value as 95, if the chroma of the pixel is less than 95, judging that the pixel is not bare soil, deleting the pixel, and otherwise, keeping the predicted pixel.
Further, the step of setting the color ratio threshold value to classify the bare soil and the cement two times and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
and taking the ratio of RGB blue to RGB red as a threshold value of 0.91, if the ratio of RGB blue to RGB red of the pixel is more than 0.91, judging that the pixel is not bare soil, deleting the pixel, and otherwise, keeping the pixel.
Further, the step of reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity comprises:
and deleting the pixel blocks with the pixel number less than 500, and reserving the pixel blocks with the pixel number more than 500.
Further, the step of dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks comprises:
outputting a prediction result, and if the number of pixel blocks in the picture is more than or equal to 3, judging that the picture has bare soil; if the number of the pixel blocks in the picture is less than 3, judging that the bare soil does not exist.
The invention also provides a bare soil identification device, which comprises:
the image input module is used for inputting an image to be detected;
the image segmentation module is used for carrying out pixel level prediction on the image shot by the construction site through the deep learning neural network pre-training model and carrying out image segmentation on the image;
the confidence coefficient filtering module is used for setting a confidence coefficient threshold value to filter the image;
the bare soil feature extraction module is used for extracting pixels identified as bare soil as an image mask and extracting features of the image by using the image mask;
the two classification modules are used for carrying out two classifications on bare soil and cement through colors and deleting pixels which are judged to be non-bare soil in the image;
the pixel block integrity filtering module is used for reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity;
and the image output module is used for dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks.
The invention also provides bare soil identification equipment, which is characterized in that: the bare soil identification device comprises:
the bare soil identification processing program realizes the steps of the bare soil identification method when being executed by the processor.
The present invention also provides a computer-readable storage medium characterized in that: the computer storage medium stores a bare soil identification processing program, and the bare soil identification processing program realizes the steps of the bare soil identification method when being executed by a processor.
The invention has the beneficial effects that:
the invention provides a bare soil identification method, a device, equipment and a computer readable storage medium, wherein the bare soil identification method carries out feature cascade by retaining the integrity of confidence, color and pixel blocks, compared with a retraining neural network model, a deep learning neural network pre-training model and a machine learning manual feature cascade method are adopted, preliminary model prediction data are provided by deep learning image segmentation, preliminary pixel level prediction is carried out on bare soil and non-bare soil areas, and a traditional machine learning manual feature cascade method is further introduced on the basis of image segmentation, so that the accuracy and recall ratio are effectively improved, the images with bare soil are successfully classified, the data do not need to be reprocessed and labeled, the development flow is simplified, the development time is short, the classification speed is high, and the algorithm performance is good. The method can be applied to different construction sites, and greatly reduces the labor cost, so that the method has the advantages of short development time, high classification speed and high accuracy.
Drawings
FIG. 1 is a flow chart of a method of bare earth identification;
FIG. 2 is a flow chart of a bare soil identification method;
FIG. 3 is a flow chart of a bare soil identification method;
FIG. 4 is a flow chart of a method of bare soil identification;
FIG. 5 is a flow chart of a method of bare soil identification;
FIG. 6 is a flow chart of a method of bare soil identification;
FIG. 7 is a flow chart of a bare soil identification method;
FIG. 8 is a detection flow diagram of a bare soil identification method;
FIG. 9 is a schematic view of a bare soil identification device;
FIG. 10 is a diagram of the effect of false soil false detection after and after cascading with a confidence of 0.9;
FIG. 11 is a diagram of the effect of false soil false detection after and after cascading with a confidence of 0.9;
FIG. 12 is a diagram of the effect of false soil false detection after cascading with a confidence level of 0.9;
FIG. 13 is a diagram of the effect of true bare soil cascading with a confidence of 0.9;
FIG. 14 is a diagram of the effect of real bare soil cascaded with a confidence of 0.9;
FIG. 15 is a diagram of the effect of true bare soil cascaded with a confidence of 0.9;
FIG. 16 is a graph showing the effect of classifying cement patterns into two categories by using the chromaticity 95 as a threshold;
FIG. 17 is a graph showing the effect of the bare soil map classified twice by using the chromaticity 95 as a threshold;
FIG. 18 is a graph of the filtering effect of the cement graph by taking 95 as a chromaticity threshold and 0.9 as a confidence threshold;
fig. 19 is a graph of the filtering effect of the bare soil by taking 95 as the chroma threshold and 0.9 as the confidence threshold.
Detailed Description
The following describes in detail the technical solutions of the embodiments of the present disclosure and how to solve the above technical problems with specific embodiments. These several specific embodiments may be combined with each other below, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present disclosure will be described below with reference to the accompanying drawings.
As shown in fig. 1 and 8, the present invention provides a bare soil identification method, which includes the following steps:
step S1, inputting an image to be detected;
s2, performing pixel level prediction on the image through a deep learning neural network pre-training model, and performing image segmentation;
specifically, image segmentation is carried out on the image through a deep learning neural network pre-training model, and a preliminary prediction result is obtained;
s3, setting a confidence threshold value to filter the image;
specifically, when the image is predicted by semantic segmentation, each pixel has confidence prediction data of each category, and the image is filtered by setting a confidence threshold value due to the fact that the difference of prediction confidence exists between real soil and false soil, so that the false detection rate of a deep learning neural network pre-training model can be reduced, and the model prediction accuracy rate is improved.
S4, extracting pixels identified as bare soil as an image mask, and performing feature extraction on the image by using the image mask;
s5, carrying out secondary classification on the bare soil and the cement through colors, and deleting pixels which are judged to be non-bare soil in the image;
specifically, as the chromaticity and the color ratio of the bare soil and the cement are obviously different, the bare soil and the cement are classified twice by setting a chromaticity threshold value and setting a color ratio threshold value, so that the misclassification is reduced.
S6, reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity;
specifically, after filtering, a small part of cement is predicted to be bare soil, but a cement prediction pixel block has the characteristic of fragmentation after being subjected to two-classification filtering, the pixel block is scattered and has no connectivity, and the integrity and the connectivity of the true bare soil prediction pixel are strong.
And S7, dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks.
The bare soil identification method carries out feature cascade through confidence degree, color and pixel block retention integrity, compared with a retraining neural network model, a deep learning neural network pre-training model and a machine learning manual feature cascade method are adopted, preliminary model prediction data are provided through deep learning image segmentation, preliminary pixel level prediction is carried out on bare soil and non-bare soil areas, a traditional machine learning manual feature cascade method is further introduced on the basis of image segmentation, accuracy and recall rate are effectively improved, pictures with bare soil are successfully classified, data do not need to be reprocessed and labeled, development procedures are simplified, development time is short, classification speed is high, and algorithm performance is good. The method can be applied to different construction sites, greatly reduces the labor cost, and has the advantages of short development time, high classification speed and high accuracy.
In step S3, as shown in fig. 2, the step of setting a confidence threshold to filter the image includes:
and S31, taking the confidence coefficient threshold value as 0.9, judging the soil to be bare soil if the pixel confidence value is greater than 0.9, and otherwise, predicting the result type according to the confidence coefficient cistron. The threshold selection method is to select the threshold with the minimum classification error rate on the sample picture according to the characteristic distribution condition by comparing results obtained by different thresholds, and essentially uses a maximum likelihood estimation method.
Specifically, fig. 10 to 12 are front and rear effect graphs in which false soil false detection is performed by cascade with a confidence of 0.9, fig. 13 to 15 are front and rear effect graphs in which true bare soil is cascaded with a confidence of 0.9, the upper half parts of fig. 10 to 15 are all unprocessed parts, and the lower half parts of fig. 10 to 15 are all effect graphs in which true bare soil is cascaded with a confidence of 0.9, it can be seen that, after false soil is filtered by a confidence threshold, regions identified as bare soil are greatly reduced, even some regions identified as bare soil are directly disappeared, and pixels of images of true soil can be retained, so that the false detection rate can be reduced and the prediction accuracy can be improved by the setting of step S31.
As shown in fig. 3, in the step S5, the step of classifying bare soil and cement into two categories by color and deleting pixels determined to be non-bare soil in the image includes:
s51, setting a chromaticity threshold value to carry out secondary classification on bare soil and cement, and deleting pixels which are judged to be non-bare soil in the image;
and S52, setting a color ratio threshold value, carrying out secondary classification on the bare soil and the cement, and deleting pixels which are judged to be non-bare soil in the image.
Specifically, the cement and the bare soil can be confirmed to have larger color difference through manual and visual comparison, so that the chromaticity is brought into an analysis range and is brought into the feature selection, and the RGB color ratio of the cement and the real bare soil has obvious difference, so that the bare soil and the cement are classified twice through the chromaticity threshold and the color ratio threshold, and the misclassification is reduced.
In step S51, the step of setting the chromaticity threshold to classify bare soil and cement into two categories and deleting the pixels determined as non-bare soil in the image includes:
as shown in fig. 4, in step S511, the chroma threshold is 95, if the chroma of the pixel is less than 95, it is determined that the pixel is not bare soil, and the pixel is deleted, otherwise the predicted pixel is retained.
Specifically, fig. 16 is an effect diagram obtained by classifying a cement diagram into two categories by using the chromaticity 95 as a threshold; FIG. 17 is a graph showing the effect of the bare soil map classified twice by using the chromaticity 95 as a threshold; FIG. 18 is a graph of the filtering effect of a cement map by taking 95 as a chroma threshold and 0.9 as a confidence threshold; fig. 19 is a graph of the filtering effect of the bare soil by taking 95 as the chroma threshold and 0.9 as the confidence threshold. It can be confirmed that the cement pixels in the image can be filtered greatly or even completely by better performing two-classification by taking the chroma threshold value as 95, and the model accuracy is improved.
As shown in fig. 5, in step S52, the step of setting the color ratio threshold to classify the bare soil and the cement into two categories and deleting the pixels determined as non-bare soil in the image includes:
and step S521, taking the ratio of RGB blue to RGB red as a threshold value, judging that the pixel is not bare soil if the ratio of RGB blue to RGB red of the pixel is more than 0.91, and deleting the pixel, otherwise, keeping the pixel.
Specifically, cement is basically gray pixels, and most of bare soil is reddish pixels, so that the ratio difference between blue and RGB (red, green, blue) of cement and RGB (red, green, blue) of the bare soil is obvious, and the ratio of RGB blue to RGB red is 0.91 serving as a threshold value, so that secondary classification can be better performed, and misclassification is reduced.
As shown in fig. 6, in step S6, the step of reserving or deleting the pixel block on the image according to the pixel block reservation integrity includes:
and S61, deleting the pixel blocks with the pixel number less than 500, and reserving the pixel blocks with the pixel number more than 500.
Specifically, the characteristic that the cement prediction pixel block is fragmented and the characteristic that the integrity and the connectivity of the real bare soil prediction pixel are strong are utilized, the pixel block with the pixel number smaller than 500 is deleted, only the pixel block with the pixel number larger than 500 is reserved, and the prediction result is further optimized.
As shown in fig. 7, in step S7, the step of dividing the predicted image into two types, i.e., bare soil and bare soil, according to the number of the complete pixel blocks includes:
step S71, outputting the prediction result, and if the number of pixel blocks in the picture is more than or equal to 3, judging that the picture has bare soil; if the number of the pixel blocks in the picture is less than 3, judging that the bare soil does not exist.
As shown in fig. 9, the present invention also provides an open soil identification device, including:
the image input module 10 is used for inputting an image to be detected;
the image segmentation module 20 is used for performing pixel level prediction on an image shot by a construction site through a deep learning neural network pre-training model, and performing image segmentation on the image;
a confidence coefficient filtering module 30, configured to set a confidence coefficient threshold to filter the image;
the bare soil feature extraction module 40 is configured to extract pixels identified as bare soil as an image mask, and perform feature extraction on the image by using the image mask;
a second classification module 50, configured to perform a second classification on the bare soil and the cement according to color, and delete a pixel determined as non-bare soil in the image;
a pixel block integrity filtering module 60, configured to reserve or delete a pixel block on the image according to the pixel block reservation integrity;
and an image output module 70, configured to divide the predicted image into two types, i.e., bare soil type and bare soil type, according to the number of the complete pixel blocks.
The device provided by the embodiment is used for inputting the image to be detected; the image segmentation method comprises the steps of predicting the pixel level of an image shot by a construction site through a deep learning neural network pre-training model, and performing image segmentation on the image; the confidence coefficient threshold value is set to filter the image; the image processing device is used for extracting pixels identified as bare soil as an image mask and extracting the features of the image by using the image mask; the device is used for carrying out secondary classification on bare soil and cement through colors, and deleting pixels which are judged to be non-bare soil in the image; the device is used for reserving or deleting the pixel blocks on the image according to the pixel block reserving integrity; the method is used for dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks.
The device carries out feature cascade through confidence coefficient, color and pixel block retention integrity, adopts a deep learning neural network pre-training model and machine learning manual feature cascade to compare with a retraining neural network model, provides preliminary model prediction data through deep learning image segmentation, carries out preliminary pixel level prediction on bare soil and non-bare soil areas, and further introduces the traditional machine learning manual feature cascade on the basis of image segmentation, thereby effectively improving accuracy and recall rate.
It should be noted that the present embodiment is an apparatus embodiment corresponding to the method embodiment described above, and the present embodiment can be implemented in cooperation with the method embodiment described above. The related technical details mentioned in the above method embodiments are still valid in this embodiment, and are not described herein again in order to reduce repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the above-described method item embodiments.
The invention also provides bare soil identification equipment, which comprises:
the bare soil identification processing program realizes the steps of the bare soil identification method when being executed by the processor.
In addition, the present invention also provides a computer readable storage medium, wherein the computer storage medium stores a bare soil identification processing program, and the bare soil identification processing program realizes the steps of the bare soil identification method when being executed by a processor.
Through the description of the foregoing embodiments, it is clear to those skilled in the art that the method of the foregoing embodiments may be implemented by software plus a necessary general hardware platform, and certainly may also be implemented by hardware, but in many cases, the former is a better implementation. Based on such understanding, the technical solution of the present invention essentially or contributing to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) as described above and includes several instructions for enabling a terminal device (such as a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.
While the embodiments of the invention have been described in terms of practical embodiments, they are not to be construed as limiting the invention, and modifications of the embodiments and combinations with other embodiments will be apparent to those skilled in the art in light of the present disclosure.
Claims (10)
1. A bare soil identification method is characterized in that: the bare soil identification method comprises the following steps:
inputting an image to be detected;
performing pixel level prediction on the image through a deep learning neural network pre-training model, and performing image segmentation;
setting a confidence threshold value to filter the image;
extracting pixels identified as bare soil as an image mask, and performing feature extraction on the image by using the image mask;
carrying out secondary classification on the bare soil and the cement through colors, and deleting pixels which are judged to be non-bare soil in the image;
reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity;
and dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks.
2. The bare soil identification method according to claim 1, wherein: the step of setting the confidence threshold value to filter the image comprises the following steps:
and (4) taking the confidence degree threshold value as 0.9, judging the soil to be bare soil if the pixel confidence degree value is greater than 0.9, and otherwise, predicting the result type according to the confidence degree forward delay.
3. The bare soil identification method according to claim 2, wherein: the step of classifying the bare soil and the cement by colors and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
setting a chromaticity threshold value to classify bare soil and cement twice, and deleting pixels which are judged to be non-bare soil in the image;
and setting a color ratio threshold value to carry out secondary classification on the bare soil and the cement, and deleting pixels which are judged to be non-bare soil in the image.
4. The bare soil identification method according to claim 3, wherein: the step of setting the chromaticity threshold value to carry out secondary classification on the bare soil and the cement, and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
and taking the chroma threshold value as 95, if the chroma of the pixel is less than 95, judging that the pixel is not bare soil, deleting the pixel, and otherwise, keeping the predicted pixel.
5. The bare soil identification method according to claim 3, wherein: the step of setting the color ratio threshold value to classify the bare soil and the cement twice and deleting the pixels which are judged to be non-bare soil in the image comprises the following steps:
and taking the ratio of RGB blue to RGB red as a threshold value, if the ratio of RGB blue to RGB red of the pixel is more than 0.91, judging that the pixel is not bare soil, deleting the pixel, and otherwise, keeping the pixel.
6. The bare soil identification method according to claim 1, wherein: the step of reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity comprises the following steps:
and deleting the pixel blocks with the pixel number less than 500, and reserving the pixel blocks with the pixel number more than 500.
7. The bare soil identification method according to claim 1, wherein: the step of dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks comprises the following steps:
outputting a prediction result, and if the number of pixel blocks in the picture is more than or equal to 3, judging that the picture has bare soil; and if the number of the pixel blocks in the picture is less than 3, judging that no bare soil exists.
8. A bare soil identification device, characterized in that: the bare soil identification device comprises:
the image input module is used for inputting an image to be detected;
the image segmentation module is used for carrying out pixel level prediction on an image shot by a construction site through a deep learning neural network pre-training model and carrying out image segmentation on the image;
the confidence coefficient filtering module is used for setting a confidence coefficient threshold value to filter the image;
the bare soil feature extraction module is used for extracting pixels identified as bare soil as an image mask and extracting features of the image by using the image mask;
the two classification modules are used for carrying out two classifications on bare soil and cement through colors and deleting pixels which are judged to be non-bare soil in the image;
the pixel block integrity filtering module is used for reserving or deleting the pixel blocks on the image according to the pixel block reservation integrity;
and the image output module is used for dividing the predicted image into two types of bare soil and bare soil according to the number of the complete pixel blocks.
9. An bare soil identification device, characterized by: the bare soil identification device comprises:
a memory, a processor, a camera and a bare soil identification processing program stored on the memory and operable on the processor, the bare soil identification processing program when executed by the processor implementing the steps of the bare soil identification method according to any one of claims 1 to 7.
10. A computer-readable storage medium characterized by: the computer storage medium has a bare soil identification processing program stored thereon, which when executed by a processor implements the steps of the bare soil identification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111601745.XA CN114299290B (en) | 2021-12-24 | 2021-12-24 | Bare soil identification method, device, equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111601745.XA CN114299290B (en) | 2021-12-24 | 2021-12-24 | Bare soil identification method, device, equipment and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114299290A CN114299290A (en) | 2022-04-08 |
| CN114299290B true CN114299290B (en) | 2023-04-07 |
Family
ID=80969511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111601745.XA Active CN114299290B (en) | 2021-12-24 | 2021-12-24 | Bare soil identification method, device, equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114299290B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600574A (en) * | 2016-08-25 | 2017-04-26 | 中国科学院遥感与数字地球研究所 | Landslide extraction method based on remote-sensing image and altitude data |
| CN113052922A (en) * | 2021-03-26 | 2021-06-29 | 重庆紫光华山智安科技有限公司 | Bare soil identification method, system, device and medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020163455A1 (en) * | 2019-02-05 | 2020-08-13 | Urugus S.A. | Automatic optimization of machine learning algorithms in the presence of target datasets |
| US11238593B2 (en) * | 2020-02-12 | 2022-02-01 | Adobe Inc. | Multi-object image parsing using neural network pipeline |
| CN111797676B (en) * | 2020-04-30 | 2022-10-28 | 南京理工大学 | High-resolution remote sensing image target on-orbit lightweight rapid detection method |
| CN111915627B (en) * | 2020-08-20 | 2021-04-16 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Semantic segmentation method, network, device and computer storage medium |
| CN112861690B (en) * | 2021-02-01 | 2024-02-02 | 武汉汉达瑞科技有限公司 | Multi-method fused remote sensing image change detection method and system |
| CN113111716B (en) * | 2021-03-15 | 2023-06-23 | 中国科学院计算机网络信息中心 | Remote sensing image semiautomatic labeling method and device based on deep learning |
| CN113378754B (en) * | 2021-06-24 | 2023-06-20 | 中国计量大学 | Bare soil monitoring method for construction site |
-
2021
- 2021-12-24 CN CN202111601745.XA patent/CN114299290B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600574A (en) * | 2016-08-25 | 2017-04-26 | 中国科学院遥感与数字地球研究所 | Landslide extraction method based on remote-sensing image and altitude data |
| CN113052922A (en) * | 2021-03-26 | 2021-06-29 | 重庆紫光华山智安科技有限公司 | Bare soil identification method, system, device and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114299290A (en) | 2022-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109151501B (en) | Video key frame extraction method and device, terminal equipment and storage medium | |
| CN109190752B (en) | Image semantic segmentation method based on global features and local features of deep learning | |
| CN112132156B (en) | Image saliency target detection method and system based on multi-depth feature fusion | |
| CN109753913B (en) | Multi-mode video semantic segmentation method with high calculation efficiency | |
| CN107784654B (en) | Image segmentation method and device and full convolution network system | |
| CN111814902A (en) | Target detection model training method, target identification method, device and medium | |
| CN109657715B (en) | Semantic segmentation method, device, equipment and medium | |
| CN112287941B (en) | License plate recognition method based on automatic character region perception | |
| CN111311475A (en) | Detection model training method and device, storage medium and computer equipment | |
| CN110020658B (en) | Salient object detection method based on multitask deep learning | |
| CN112668522B (en) | Human body key point and human body mask joint detection network and method | |
| CN113034506B (en) | Remote sensing image semantic segmentation method and device, computer equipment and storage medium | |
| CN111160356A (en) | Image segmentation and classification method and device | |
| CN113052170A (en) | Small target license plate recognition method under unconstrained scene | |
| CN109977738B (en) | Video scene segmentation judging method, intelligent terminal and storage medium | |
| CN107871315B (en) | Video image motion detection method and device | |
| CN113887649A (en) | Target detection method based on fusion of deep-layer features and shallow-layer features | |
| CN114299290B (en) | Bare soil identification method, device, equipment and computer readable storage medium | |
| CN112348762A (en) | Single image rain removing method for generating confrontation network based on multi-scale fusion | |
| CN112785610A (en) | Lane line semantic segmentation method fusing low-level features | |
| Sridevi et al. | Vehicle identification based on the model | |
| CN114449362B (en) | Video cover selection method, device, equipment and storage medium | |
| CN107341456B (en) | Weather sunny and cloudy classification method based on single outdoor color image | |
| CN112991239B (en) | Image reverse recovery method based on deep learning | |
| Hanbury | How do superpixels affect image segmentation? |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |