Bare soil identification method and device, equipment and medium

Abstract

The embodiment of the invention discloses a bare soil identification method and device, equipment and a medium. The method comprises the following steps: dividing a to-be-recognized image of a to-be-recognized area into a plurality of candidate areas; determining a feature consistency judgment result of each candidate region according to the image features in the candidate regions, and determining whether the candidate regions are single feature regions or mixed feature regions according to the feature consistency judgment results of the candidate regions; determining a first bare soil identification result of the single feature region based on a pre-trained classifier; bare soil information in the mixed feature region is determined according to the bare soil feature information to serve as a second bare soil recognition result; and determining a bare soil identification result of the to-be-identified area according to the first bare soil identification result and the second bare soil identification result. According to the technical scheme, region consistency division is carried out on the image, bare soil recognition is carried out on the divided single feature region and the mixed feature region in different modes, and the accuracy of bare soil recognition is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a bare soil identification method, device, equipment and medium. Background Art

[0002] Today's society pays more and more attention to air quality and environmental pollution. Among them, dust-proof covering of bare soil at construction sites has become a focus of attention. Currently, there are two main bare soil covering solutions: 1. Directly obtain the bare soil area through deep learning semantic segmentation method and calculate the bare soil coverage rate; 2. Obtain the number of pixels of bare soil or covering network through traditional visual algorithm and calculate the bare soil coverage rate.

[0003] However, all of the above solutions have certain limitations. In the first solution, the semantic segmentation model has a greater risk of missing detection when the bare soil areas are scattered and varied in shape in actual construction sites. In the second solution, the error is large when the coverage network is sparse, and the threshold needs to be reset when the bare soil color is different in different scenes. Therefore, the accuracy of bare soil recognition in the above two solutions is poor and cannot meet actual detection needs. Summary of the invention

[0004] The present invention provides a bare soil identification method, device, equipment and medium, which divide an image into regions with consistency and use different methods to identify bare soil in the divided single feature regions and mixed feature regions, thereby effectively improving the accuracy of bare soil identification.

[0005] According to one aspect of the present invention, a bare soil identification method is provided, the method comprising:

[0006] Dividing the image to be identified in the area to be identified into a plurality of candidate areas;

[0007] Determine the feature consistency judgment result of each candidate region according to the image features in the candidate region, and determine whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment result of the candidate region;

[0008] Determining a first bare soil identification result of the single characteristic area based on a pre-trained classifier;

[0009] Determining bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result;

[0010] The bare soil identification result of the area to be identified is determined according to the first bare soil identification result and the second bare soil identification result.

[0011] According to another aspect of the present invention, there is provided a bare soil identification device, comprising:

[0012] A candidate region division module is used to divide the image to be identified into a plurality of candidate regions;

[0013] A feature consistency judgment module, used to determine the feature consistency judgment results of each candidate region according to the image features in the candidate region, and determine whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment results of the candidate region;

[0014] A first bare soil identification result determination module, used to determine a first bare soil identification result of the single feature area based on a pre-trained classifier;

[0015] A second bare soil identification result determination module, configured to determine bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result;

[0016] The bare soil identification result determination module is used to determine the bare soil identification result of the area to be identified according to the first bare soil identification result and the second bare soil identification result.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0018] at least one processor; and

[0019] a memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the bare soil identification method described in any embodiment of the present invention.

[0021] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the bare soil identification method described in any embodiment of the present invention when executed.

[0022] The technical solution of the embodiment of the present invention divides the image to be identified in the area to be identified into multiple candidate areas; determines the feature consistency judgment results of each candidate area according to the image features in the candidate area, and determines the candidate area as a single feature area or a mixed feature area according to the feature consistency judgment results of the candidate area; determines the first bare soil recognition result of the single feature area based on a pre-trained classifier; determines the bare soil information in the mixed feature area according to the bare soil feature information as the second bare soil recognition result; determines the bare soil recognition result of the area to be identified according to the first bare soil recognition result and the second bare soil recognition result. This technical solution divides the image into regions according to regional consistency and uses different methods to perform bare soil recognition on the single feature area and the mixed feature area after division, which effectively improves the bare soil recognition accuracy.

[0023] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 is a flow chart of a bare soil identification method provided according to Embodiment 1 of the present invention;

[0026] Figure 2 is a schematic diagram of determining feature similarity of candidate regions according to Embodiment 1 of the present invention;

[0027] Figure 3 is a flow chart of a bare soil identification method provided according to Embodiment 2 of the present invention;

[0028] Figure 4 is a schematic diagram of a region expansion provided according to Embodiment 2 of the present invention;

[0029] Figure 5 is a schematic structural diagram of a bare soil identification device provided according to Embodiment 3 of the present invention;

[0030] Figure 6 It is a structural schematic diagram of an electronic device for implementing a bare soil identification method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first", "second", "target", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] Embodiment 1

[0034] Figure 1 This is a flowchart of a bare soil identification method provided in the first embodiment of the present invention. This embodiment is applicable to the situation of accurately identifying bare soil. The method can be executed by a bare soil identification device. The bare soil identification device can be implemented in the form of hardware and / or software. The bare soil identification device can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:

[0035] S110, dividing the image to be recognized in the area to be recognized into a plurality of candidate areas.

[0036] Among them, the area to be identified may refer to the target area waiting for bare soil identification, which can be set specifically according to actual detection needs. The image to be identified may refer to a photographed image of the area to be identified. The candidate area may refer to any image sub-area obtained by dividing the image to be identified. It should be noted that this embodiment does not specifically limit the number and size of candidate areas, which can be flexibly set according to actual application scenarios. Exemplarily, the image to be identified can be divided into N×M fixed-size areas as candidate areas, where N and M represent the number of blocks of the image to be identified in the width and height directions, respectively.

[0037] S120, determining feature consistency judgment results of each candidate region according to the image features in the candidate region, and determining whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment results of the candidate region.

[0038] Among them, the image features may include color features and / or texture features. The feature consistency judgment result may include having feature consistency or not having feature consistency. A single feature area may refer to an area where the image features are consistent (belonging to the same category) within the candidate area, such as a candidate area that only contains bare soil or covered nets. A mixed feature area may refer to an area where the image features are inconsistent (mixed with different categories) within the candidate area, such as a candidate area where bare soil and covered nets each occupy 1 / 2.

[0039] In this embodiment, after the image to be identified in the area to be identified is divided into multiple candidate areas, the feature consistency judgment results of each candidate area can be determined respectively according to the image features in the candidate area. Optionally, the feature consistency judgment results of each candidate area are determined respectively according to the image features in the candidate area, including: dividing the candidate area into multiple sub-areas; respectively determining the image feature vector of each sub-area; wherein the image feature vector includes at least color dimension features and / or texture dimension features; determining the feature similarity of the candidate area according to the image feature vector of the sub-area; if the feature similarity is greater than a preset similarity threshold, determining the feature consistency judgment result of the candidate area as having feature consistency; otherwise, determining the feature consistency judgment result of the candidate area as not having feature consistency.

[0040] Among them, the sub-region may refer to any sub-region in the candidate region obtained by dividing the candidate region. It should be noted that this embodiment does not specifically limit the number and size of the sub-regions, which can be set according to the actual application scenario. The preset similarity threshold may refer to a pre-set feature similarity reference value, which can be set according to actual needs. For example, the preset similarity threshold can be set to 0.5.

[0041] In this embodiment, when determining the feature consistency judgment results of each candidate region according to the image features in the candidate region, it is first necessary to divide the candidate region into multiple sub-regions, and determine the image feature vectors of each sub-region respectively. Among them, the image feature vector includes at least color dimension features and / or texture dimension features. Exemplarily, the candidate region can be divided into 9 sub-regions, and then the image feature vectors of the 9 sub-regions are determined respectively.

[0042] Specifically, for color dimension features, the color features can be described by calculating the color moment (including first-order color moment and second-order color moment) of each sub-region. Among them, the first-order color moment is used to reflect the brightness of the image, and the second-order color moment is used to reflect the color distribution range of the image. The calculation method of the first-order color moment is expressed as: The calculation method of the second-order color moment is expressed as: Among them, μ i is the first-order color moment, σ i is the second-order color moment, i is the color channel (r, g, b), j is the pixel position, P ij represents the color value at the i color channel and j pixel position. Therefore, the color dimension feature can be expressed as (μ r , μ g , μ b , σ r , σ g , σ b ).

[0043] For texture dimension features, it can be expressed as approximate grayscale differential entropy. Specifically, first convert the RGB image into a grayscale image, then calculate the difference between (x, y) and (x+Δx, y+Δy) based on the grayscale value of the point (x, y) in the image and its nearby point (x+Δx, y+Δy), generate a histogram, and then calculate the entropy based on the difference value. The entropy calculation method can be expressed as: e=∑p m log2[p m ], where m represents the difference value, p m represents the probability of occurrence of the differential value m. Therefore, the texture dimension feature can be represented as e.

[0044] In this embodiment, if the image feature vector includes both color dimension features and texture dimension features, the color dimension features and texture dimension features can be fused to obtain a fused feature vector (μ r , μ g , μ b , σ r , σ g , σ b , e). Furthermore, the feature vectors of the same dimension can be weighted using the same weight, and the weighted feature vector is (λ1(μ r , μ g , μ b ),λ2(σ r , σ g , σ b ), λ3e), where λ1+λ2+λ3=1. Thus, the weighted feature vector can be used as the image feature vector.

[0045] It should be noted that the value of each weight determines the importance of each dimensional feature in the feature consistency judgment. Due to the subjectivity of human judgment, 5000 sets of real scenes (including background targets that appear in bare soil scenes such as bare soil, covered nets, sky, hard road surface, etc.) were collected after data sampling, and the weights were roughly estimated by the method used by real data. Specifically, first add labels to the 5000 sets of real scenes, mark the same category with the same label, and mark different categories with different labels. Then, based on the loss function, the distance between the same target features is made as close as possible, and the distance between different target features is made as far away as possible, and the optimal weight parameters can be obtained by artificial setting or by using any search method (such as neural network search method).

[0046] In this embodiment, after the image feature vectors of each sub-region are determined respectively, the feature similarity of the candidate region can be determined according to the image feature vectors of the sub-region. Exemplarily, the feature similarity of the candidate region can be determined by calculating the similarity between the image feature vectors of each sub-region. For example, assuming that the candidate region is divided into 9 sub-regions, 9 image feature vectors can be obtained according to the above process (each sub-region corresponds to 1 image feature vector). Then the similarity is calculated once between every two image feature vectors to obtain a total of 36 similarities, and the minimum value of the 36 similarities is determined as the feature similarity of the candidate region.

[0047] In this embodiment, optionally, the feature similarity of the candidate region is determined based on the image feature vector of the sub-region, including: inputting the image feature vector of the sub-region into a pre-trained consistency neural network discriminator, fusing and extracting the image feature vector through the network, and obtaining an intermediate feature vector corresponding to the image feature vector; respectively determining the candidate feature similarities between the two intermediate feature vectors; and taking the minimum value among the candidate feature similarities as the feature similarity of the candidate region.

[0048] Among them, the consistency neural network discriminator can be used to discriminate the feature consistency of the candidate region. Exemplarily, the consistency neural network discriminator can be a BP (Back Propagation) neural network discriminator. The intermediate feature vector can refer to a new feature vector obtained after the image feature vector is fused and extracted by the consistency neural network. The candidate feature similarity can refer to the similarity between any two intermediate feature vectors.

[0049] In this embodiment, the accuracy of directly using the image feature vector of the sub-region to judge the feature similarity is usually less than 90%. In order to improve the accuracy of the feature similarity judgment, a consistent neural network is used here to further extract and fuse the image feature vector, and the feature similarity is judged based on the intermediate feature vector obtained after extraction and fusion. Figure 2A schematic diagram of determining feature similarity of candidate regions provided in the first embodiment of the present invention. Figure 2 As shown in the figure, assume that the candidate region is divided into 9 sub-regions, and each sub-region corresponds to 1 image feature vector. First, the image feature vectors 1-9 of the 9 sub-regions are input into the pre-trained BP neural network discriminator. After the BP neural network fuses and extracts the 9 image feature vectors, the intermediate feature vector xz corresponding to the image feature vectors 1-9 can be obtained. Then, the similarity between each two intermediate feature vectors is calculated as the candidate feature similarity, and the minimum value of the candidate feature similarity is determined as the feature similarity of the candidate region.

[0050] Through such a setting, this solution uses a consistent neural network to extract and fuse image feature vectors, and performs feature similarity judgment based on the intermediate feature vectors obtained after extraction and fusion, which can improve the accuracy of feature similarity judgment.

[0051] After determining the feature similarity of the candidate region, the feature similarity can be compared with a preset similarity threshold, and the feature consistency judgment result of the candidate region can be determined based on the comparison result. Specifically, if the feature similarity is greater than the preset similarity threshold, it indicates that the feature similarity of the candidate region is large, and the feature consistency judgment result of the candidate region can be determined as having feature consistency; otherwise, it indicates that the feature similarity of the candidate region is small, and the feature consistency judgment result of the candidate region can be determined as not having feature consistency. Accordingly, determining whether the candidate region is a single feature region or a mixed feature region based on the feature consistency judgment result of the candidate region includes: if the feature consistency judgment result of the candidate region is that it has feature consistency, determining the candidate region is a single feature region; if the feature consistency judgment result of the candidate region is that it does not have feature consistency, determining the candidate region is a mixed feature region.

[0052] S130, determining a first bare soil recognition result of a single feature area based on a pre-trained classifier.

[0053] Wherein, the pre-trained classifier can be used to classify the single feature area to determine whether the single feature area is bare soil according to the classification result. Exemplarily, the classifier can be CNN (Convolutional Neural Network) or SVM (Support Vector Machines). The first bare soil recognition result can refer to the bare soil recognition result of the single feature area.

[0054] In this embodiment, after determining that the candidate area is a single feature area or a mixed feature area, the first bare soil recognition result of the single feature area can be determined based on a pre-trained classifier. Among them, the bare soil recognition rate of the classifier for the single feature area is much higher than that of the mixed feature area. Therefore, for the bare soil recognition of the single feature area, only the classification and discrimination of the classifier is required to obtain a better discrimination effect. Exemplarily, each single feature area can be input into the pre-trained classifier one by one to obtain the first bare soil recognition result of each single feature area. For the bare soil recognition of the mixed feature area, the effect of using the classifier to classify and discriminate the bare soil is general, and it is impossible to discriminate the actual proportion of bare soil, so it is necessary to use other methods to identify the bare soil of the mixed feature area.

[0055] In this embodiment, optionally, a first bare soil identification result of a single feature area is determined based on a pre-trained classifier, including: determining an image feature vector of the single feature area; determining the regional similarity of two single feature areas respectively according to the image feature vector of the single feature area; if the regional similarity is greater than a preset similarity threshold, determining that the regional labels of the two single feature areas are the same; otherwise, determining that the regional labels of the two single feature areas are different until the regional label of each single feature area is determined; inputting the single feature area with the same regional label into the classifier to obtain a first bare soil identification result corresponding to the regional label.

[0056] In this embodiment, when determining the first bare soil identification result of a single feature area, the image feature vector of the single feature area is first determined. Among them, the method for determining the image feature vector of the single feature area can refer to the specific process of determining the image feature vector of the sub-area of ​​the candidate area mentioned above, which will not be repeated here. Then, the regional similarity of the two single feature areas is calculated respectively according to the image feature vector of the single feature area, wherein the regional similarity can be the similarity between any two single feature areas. If the regional similarity is greater than the preset similarity threshold, it indicates that the two single feature areas are highly similar. At this time, it can be determined that the regional labels of the two single feature areas are the same, otherwise it can be determined that the regional labels of the two single feature areas are different, until the regional label of each single feature area is determined. Among them, the regional labels can include bare soil, covering nets and vegetation, etc., which can be determined specifically according to the actual application scenario.

[0057] It is understandable that single feature regions with the same region label belong to the same category, and single feature regions with different region labels belong to different categories. Therefore, whether single feature regions belong to the same category can be determined based on whether the region labels are the same, and single feature regions with the same region labels can be placed in the same feature region set. In other words, there are as many feature region sets as there are categories of region labels, i.e., the number of region labels is consistent with the number of feature region sets.

[0058] After determining the region label of each single feature region, the single feature regions with the same region label can be input into the classifier as a feature region set, so as to obtain the first bare soil identification result corresponding to the region label, that is, the single feature regions with the same region label have the same first bare soil identification result. It should be noted that, compared with the method of inputting the single feature regions one by one into the pre-trained classifier to determine the first bare soil identification result of the single feature region, inputting the single feature regions with the same region label into the classifier to determine the first bare soil identification result corresponding to the region label can effectively improve the efficiency of determining the first bare soil identification result.

[0059] Through such a setting, the present solution inputs a single feature region with the same region label into the classifier to determine the first bare soil identification result corresponding to the region label, which can effectively improve the efficiency of determining the first bare soil identification result.

[0060] S140, determining bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result.

[0061] Among them, the bare soil feature information can be predetermined or obtained based on the first bare soil recognition result. Exemplarily, a large number of bare soil images can be collected in advance, and the bare soil images can be divided into regions to obtain bare soil regions. In actual scenes, even if two regions belong to bare soil regions, there are obvious differences between their corresponding feature vectors. Therefore, different bare soil category labels can be pre-set as label_1, label_2, ..., label_N according to actual application requirements. After obtaining the bare soil region, each bare soil region can be input into a pre-trained classifier to obtain the bare soil category label corresponding to each bare soil region, and the bare soil regions with the same bare soil category label are placed in a bare soil set. It can be understood that there are several categories of pre-set bare soil category labels, and there are several corresponding bare soil sets, that is, the number of bare soil category labels is consistent with the number of bare soil sets. In actual scenes, the pixel differences of bare soil images are mainly reflected in color and brightness features, so the bare soil feature information can be determined based on the color features and brightness features of the bare soil image.

[0062] Specifically, for each bare soil set, the bare soil area in the bare soil set is first converted from RGB (red-green-blue) color form to HSV (hue-saturation-brightness) color form, and the features of H and V are extracted from it. Among them, there is a nonlinear correspondence between H and V, and it is impossible to directly use a histogram to distinguish. The range of H can be divided into a preset number of levels (such as 5) from small to large. In each bare soil set, the distribution range of V is statistically calculated corresponding to the level division interval of H, thereby obtaining the mapping relationship between H and V, which can be used as bare soil feature information.

[0063] It should be noted that the preset number can be set according to actual needs, and this embodiment does not specifically limit this. In addition, since different bare soil sets correspond to different bare soil category labels, the levels of H and the distribution ranges of V corresponding to different bare soil sets are different, that is, the mapping relationship between H and V (that is, bare soil feature information) corresponding to different bare soil sets is different. Since the bare soil feature information is predetermined, when the bare soil feature information is needed, it can be directly obtained by searching, thereby improving the efficiency of determining the bare soil feature information.

[0064] Exemplarily, assuming that the preset number is 5, Table 1 shows the bare soil feature information of the bare soil category label label_1.

[0065] Table 1 Bare soil feature information of bare soil category label label_1

[0066] H Level H1(8-16) H2(16-24) H3(24-32) H4(32-40) H5(40-50) The distribution range of V 54-75 65-88 75-92 85-91 78-95

[0067] Among them, H1-H5 represent 5 different H levels.

[0068] Exemplarily, assuming that the preset number is 5, Table 2 shows the bare soil feature information of the bare soil category label label_2.

[0069] Table 2 Bare soil feature information of bare soil category label label_2

[0070] H Level H1(13-18) H2(18-24) H3(24-30) H4(30-36) H5(36-45) The distribution range of V 140-162 145-172 142-165 150-170 150-160

[0071] The bare soil information may include bare soil position information, and the bare soil position information may be used to characterize the pixel position where the bare soil is located in the mixed feature area. The second bare soil recognition result may refer to the bare soil recognition result of the mixed feature area.

[0072] In this embodiment, optionally, bare soil information in the mixed feature area is determined based on the bare soil feature information as the second bare soil identification result, including: determining a single bare soil area in the single feature area based on the first bare soil identification result; determining bare soil feature information based on bare soil color features and bare soil brightness features in the single bare soil area; dividing the mixed feature area into multiple candidate sub-areas; if the sub-area color features and sub-area brightness features in the candidate sub-areas in the mixed feature area successfully match the bare soil color features and bare soil brightness features, then determining that the candidate sub-area is a mixed bare soil area in the mixed feature area, and determining the second bare soil identification result based on the mixed bare soil area.

[0073] Among them, a single bare soil area may refer to a bare soil area in a single feature area. A candidate sub-area may refer to any sub-area obtained by dividing a mixed feature area. An exemplary candidate sub-area may include only one pixel, or may be set to an area of ​​a preset size (such as 2×2). A sub-area color feature may refer to a color feature of a candidate sub-area. A sub-area brightness feature may refer to a brightness feature of a candidate sub-area. A bare soil color feature may refer to a color feature of a single bare soil area. A bare soil brightness feature may refer to a brightness feature of a single bare soil area. A mixed bare soil area may refer to a bare soil area in a mixed feature area.

[0074] In this embodiment, the bare soil feature information can be determined based on the first bare soil recognition result. Specifically, firstly, a single bare soil region in a single feature region is determined according to the first bare soil recognition result, and then the single bare soil region is converted from the RGB color form to the HSV color form, and the features of H and V are extracted therefrom as the bare soil color feature and the bare soil brightness feature, respectively. Then, the single bare soil region is input into a pre-trained classifier, and the bare soil category label corresponding to each single bare soil region is determined based on different pre-set bare soil category labels, and the single bare soil regions with the same bare soil category label are placed in a single bare soil set. Then, for the single bare soil region in each single bare soil set, the range of H is divided into a preset number of levels from small to large, and the distribution range of V is statistically calculated corresponding to the level division range of H, thereby obtaining a mapping relationship between H and V, and the mapping relationship can be determined as bare soil feature information. Compared with the method of pre-determining bare soil feature information based on a large number of bare soil images collected in advance, the method of determining bare soil feature information based on the first bare soil recognition result is more targeted to the bare soil region, so the accuracy of determining the bare soil feature information can be improved.

[0075] After determining the bare soil feature information, the bare soil information in the mixed feature area can be determined according to the bare soil feature information as the second bare soil identification result. Specifically, the mixed feature area can be divided into multiple candidate sub-areas first, and then the sub-area color features and sub-area brightness features in the candidate sub-areas in the mixed feature area can be judged respectively, and whether each candidate sub-area is bare soil can be determined according to the judgment result. Among them, this embodiment does not specifically limit the number and division method of the mixed feature area, and can be set according to actual needs. When the candidate sub-area is a pixel point, it can be judged in a pixel-by-pixel traversal manner. At this time, the judgment efficiency is low, but the judgment accuracy is high; when the candidate sub-area is an area of ​​a preset size, it can be judged in a region-by-region traversal manner. At this time, the judgment efficiency is high, but the judgment accuracy is low. Therefore, this embodiment does not specifically limit the size of the candidate sub-area, and it can be reasonably set according to actual needs.

[0076] If the sub-region color features and sub-region brightness features in the candidate sub-region in the mixed feature region simultaneously satisfy a certain set of H and V mapping relationships in the bare soil feature information, it indicates that the sub-region color features and sub-region brightness features in the candidate sub-region in the mixed feature region successfully match the bare soil color features and bare soil brightness features. At this time, the candidate sub-region can be determined as a mixed bare soil region in the mixed feature region, and the second bare soil recognition result can be determined based on the mixed bare soil region. Among them, the implementation method of determining the second bare soil recognition result based on the mixed bare soil region can refer to the detailed description of determining the first bare soil recognition result of a single feature region based on the pre-trained classifier, which will not be repeated here.

[0077] Through such a setting, the present scheme makes the method of determining bare soil characteristic information based on the first bare soil identification result more targeted to the bare soil area, improves the accuracy of determining the bare soil characteristic information, so as to determine the bare soil information in the mixed characteristic area based on the bare soil characteristic information, thereby improving the accuracy of the second bare soil identification result.

[0078] S150, determining a bare soil identification result of the area to be identified according to the first bare soil identification result and the second bare soil identification result.

[0079] In this embodiment, after the first bare soil recognition result and the second bare soil recognition result are determined respectively, the first bare soil recognition result and the second bare soil recognition result can be integrated to obtain the bare soil recognition result of the area to be recognized. Compared with the prior art, the present technical solution can obtain a specific and accurate bare soil area based on the first bare soil recognition result of the single feature area in the image to be recognized and the second bare soil recognition result of the mixed feature area, effectively avoiding the problem of missing detection of bare soil areas in the image in the prior art, thereby improving the accuracy of bare soil recognition.

[0080] The technical solution of the embodiment of the present invention divides the image to be identified in the area to be identified into multiple candidate areas; determines the feature consistency judgment results of each candidate area according to the image features in the candidate area, and determines the candidate area as a single feature area or a mixed feature area according to the feature consistency judgment results of the candidate area; determines the first bare soil recognition result of the single feature area based on a pre-trained classifier; determines the bare soil information in the mixed feature area according to the bare soil feature information as the second bare soil recognition result; determines the bare soil recognition result of the area to be identified according to the first bare soil recognition result and the second bare soil recognition result. This technical solution divides the image into regions according to regional consistency and uses different methods to perform bare soil recognition on the single feature area and the mixed feature area after division, which effectively improves the bare soil recognition accuracy.

[0081] Embodiment 2

[0082] Figure 3 A flowchart of a bare soil identification method provided in the second embodiment of the present invention, this embodiment is optimized based on the above embodiment. The specific optimization is: before determining the first bare soil identification result of a single feature area based on a pre-trained classifier, the method also includes: determining the boundary of the single feature area and the mixed feature area; extending along the boundary to the mixed feature area to determine the image features of the extended area; if the image features of the extended area are more similar to the image features of the single feature area adjacent to the boundary than a preset threshold, the extended area is determined to be a single feature area; otherwise, the extension is stopped; and the boundary of the single feature area and the mixed feature area is updated according to the extended position; the updated mixed feature area is determined according to the updated boundary of the single feature area and the mixed feature area; the updated mixed feature area is used as the image to be identified, and the image to be identified is continued to be divided into multiple candidate areas, and the feature consistency judgment results of each candidate area are determined according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is no single feature area, then the final mixed feature area and the final single feature area are determined.

[0083] like Figure 3 As shown, the method of this embodiment specifically includes the following steps:

[0084] S210, dividing the image to be recognized in the area to be recognized into a plurality of candidate areas.

[0085] S220, determining feature consistency judgment results of each candidate region according to the image features in the candidate region, and determining whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment results of the candidate region.

[0086] S230, determining a boundary between the single feature region and the mixed feature region, extending the mixed feature region along the boundary, and determining image features of the extended region.

[0087] It should be noted that in the entire bare soil recognition process, the main time-consuming step is to traverse and judge the sub-region color features and sub-region brightness features in the candidate sub-regions in the mixed feature region. Therefore, the region expansion method can be used to expand the proportion of single feature regions in the image to be recognized and reduce the proportion of mixed feature regions in the image to be recognized, thereby reducing computing power and improving bare soil recognition efficiency and accuracy.

[0088] Among them, the extended area may refer to a new area obtained by extending along the boundary of the single feature area and the mixed feature area to the mixed feature area. Furthermore, in order to avoid the influence of the area size on subsequent calculations, the new area obtained after the area expansion can be extended to the single feature area adjacent to the boundary, so that the sum of the size of the new area obtained by the area expansion and the area obtained by the area extension is the same as the area size of the single feature area, and the sum of the new area obtained by the area expansion and the area obtained by the area extension is used as the extended area. Exemplarily, assuming that the size of the single feature area is 45×45, and the size of the area extended along the boundary to the mixed feature area is 1×45, then the expanded area of ​​1×45 can be directly used as the extended area, or the sum of the expanded area of ​​1×45 and the area of ​​44×45 in the single feature area adjacent to the boundary can be used as the extended area. The image feature may refer to the mean of the three channels of R, G, and B in the extended area.

[0089] S240, determining whether the similarity between the image feature of the extended area and the image feature of the single feature area adjacent to the boundary is greater than a preset threshold, if so, executing S250, otherwise executing S260-S280.

[0090] The preset threshold may refer to a preset image feature similarity reference value, which may be set according to actual needs. For example, the preset threshold may be set to 30. The image feature similarity may refer to the similarity between the image feature of the extended area and the image feature of the single feature area adjacent to the boundary.

[0091] S250: Determine that the extended area is a single characteristic area.

[0092] If the similarity between the image features of the extended area and the image features of the single feature area adjacent to the boundary is greater than a preset threshold, it indicates that the image features have a large similarity, and the extended area can be directly determined as a single feature area.

[0093] S260, stop the expansion, and update the boundaries of the single feature region and the mixed feature region according to the expansion position.

[0094] Among them, if the similarity between the image features of the expanded area and the image features of the single feature area adjacent to the boundary is less than or equal to the preset threshold, it indicates that the image feature similarity is small. At this time, it is necessary to stop the expansion and update the boundaries of the single feature area and the mixed feature area according to the expanded position.

[0095] S270, determining an updated mixed feature region according to the boundaries of the updated single feature region and the mixed feature region.

[0096] The updated mixed feature region may refer to a new mixed feature region obtained after region expansion. Specifically, a closed contour may be formed according to the boundary of the updated single feature region and the mixed feature region and the image boundary of the image to be identified which is perpendicular to the boundary direction, and the region surrounded by the closed contour is used as the updated mixed feature region. It is understandable that the updated mixed feature region and the updated single feature region are combined into the image to be identified.

[0097] S280, taking the updated mixed feature area as the image to be identified, continuing to divide the image to be identified into multiple candidate areas, determining the feature consistency judgment result of each candidate area according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is that there is no single feature area, then determining the final mixed feature area and the final single feature area.

[0098] In this embodiment, in order to further expand the proportion of single feature areas in the image to be identified and improve the efficiency and accuracy of bare soil identification, after determining the updated mixed feature area, the updated mixed feature area can be used as the image to be identified, and the image to be identified can be divided into multiple candidate areas. The feature consistency judgment results of each candidate area are determined according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is that there is no single feature area. At this time, the final mixed feature area and the final single feature area can be determined.

[0099] S290, determining a first bare soil recognition result of a single feature area based on a pre-trained classifier.

[0100] S2100, determining bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result.

[0101] S2110, determining a bare soil identification result of the area to be identified according to the first bare soil identification result and the second bare soil identification result.

[0102] Among them, the specific implementation method of S290-S2110 can refer to the detailed description of S130-S150 in Example 1, and will not be repeated here.

[0103] The technical solution of the embodiment of the present invention, before determining the first bare soil recognition result of the single feature area based on the pre-trained classifier, determines the boundary of the single feature area and the mixed feature area; expands along the boundary to the mixed feature area to determine the image feature of the extended area; if the image feature of the extended area is greater than the image feature similarity of the single feature area adjacent to the boundary is greater than a preset threshold, the extended area is determined to be a single feature area; otherwise, the expansion is stopped; and the boundary of the single feature area and the mixed feature area is updated according to the extended position; the updated mixed feature area is determined according to the updated boundary of the single feature area and the mixed feature area; the updated mixed feature area is used as the image to be recognized, and the image to be recognized is continued to be divided into multiple candidate areas, and the feature consistency judgment result of each candidate area is determined according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is no single feature area, then the final mixed feature area and the final single feature area are determined. This technical solution, on the basis of performing regional consistency division on the image and using different methods to perform bare soil recognition on the single feature area and the mixed feature area after division, adopts the regional expansion method to expand the proportion of the single feature area in the image to be recognized, thereby improving the efficiency of bare soil recognition, and further improving the accuracy of bare soil recognition.

[0104] In this embodiment, optionally, before expanding along the boundary to the mixed feature area, the method also includes: determining an image feature vector of a single feature area; determining the regional similarity of two adjacent single feature areas respectively according to the image feature vectors of the single feature area; merging adjacent single feature areas according to the regional similarity to obtain a merged single feature area; accordingly, if the image feature of the extended area has a similarity with the image feature of the single feature area adjacent to the boundary that is greater than a preset threshold, then determining that the extended area is a single feature area, including: if the image feature of the extended area has a similarity with the image feature of the merged single feature area adjacent to the boundary that is greater than a preset threshold, then determining that the extended area is a single feature area.

[0105] In this embodiment, when performing region expansion, if the expansion is performed along the boundary to the mixed feature region by traversing each single feature region, the expansion efficiency is low. Therefore, in order to improve the efficiency of region expansion, before expanding along the boundary to the mixed feature region, the image feature vector of the single feature region can be determined, and the region similarity of two adjacent single feature regions can be determined respectively according to the image feature vector of the single feature region, and then the adjacent single feature regions can be merged according to the region similarity to obtain a merged single feature region. Thus, the region expansion can be performed by traversing the merged single feature regions to obtain the extended region, thereby improving the efficiency of region expansion. Further, the image feature similarity of the extended region and the merged single feature region adjacent to the boundary is determined, and if the feature image similarity is greater than a preset threshold, the extended region can be determined as a single feature region.

[0106] However, the method of traversing and merging single feature areas to expand the region may result in poor region expansion accuracy. Therefore, in order to balance the efficiency and accuracy of region expansion, after obtaining the merged single feature area, you can still use the method of traversing each single feature area to expand the extended area. However, at this time, it is necessary to determine whether the extended area is a single feature area based on the similarity between the image features of the extended area and the image features of the merged single feature area adjacent to the boundary and a preset threshold. Figure 4 A schematic diagram of area expansion provided in Embodiment 2 of the present invention.

[0107] Through such a setting, the present solution determines whether the extended region is a single feature region based on merging single feature regions, which can effectively balance the efficiency and accuracy of region expansion.

[0108] Embodiment 3

[0109] Figure 5 This is a schematic diagram of the structure of a bare soil identification device provided in the third embodiment of the present invention. The device can execute the bare soil identification method provided in any embodiment of the present invention and has the corresponding functional modules and beneficial effects of the execution method. Figure 5 As shown, the device comprises:

[0110] A candidate region division module 310 is used to divide the image to be identified into a plurality of candidate regions;

[0111] A feature consistency judgment module 320 is used to determine the feature consistency judgment results of each candidate region according to the image features in the candidate region, and determine whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment results of the candidate region;

[0112] A first bare soil identification result determination module 330, configured to determine a first bare soil identification result of the single feature area based on a pre-trained classifier;

[0113] A second bare soil identification result determination module 340 is used to determine the bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result;

[0114] The bare soil identification result determination module 350 is used to determine the bare soil identification result of the to-be-identified area according to the first bare soil identification result and the second bare soil identification result.

[0115] Optionally, the feature consistency judgment module 320 includes:

[0116] A sub-region division unit, used for dividing the candidate region into a plurality of sub-regions;

[0117] An image feature vector determination unit, used to determine the image feature vector of each sub-region respectively; wherein the image feature vector at least includes a color dimension feature and / or a texture dimension feature;

[0118] A feature similarity determination unit, configured to determine the feature similarity of the candidate region according to the image feature vector of the sub-region;

[0119] A first judgment result determination unit, configured to determine that a feature consistency judgment result of the candidate region is feature consistency if the feature similarity is greater than a preset similarity threshold;

[0120] A second judgment result determination unit is used to determine that the feature consistency judgment result of the candidate region is not feature consistent;

[0121] Accordingly, the feature consistency judgment module 320 is used to:

[0122] If the feature consistency judgment result of the candidate region is that the candidate region has feature consistency, the candidate region is determined to be a single feature region;

[0123] If the feature consistency judgment result of the candidate region is that there is no feature consistency, the candidate region is determined to be a mixed feature region.

[0124] Optionally, the feature similarity determination unit is specifically used to:

[0125] Inputting the image feature vector of the sub-region into a pre-trained consistency neural network discriminator, fusing and extracting the image feature vector through the network, and obtaining an intermediate feature vector corresponding to the image feature vector;

[0126] Determine the candidate feature similarity between two intermediate feature vectors respectively;

[0127] The minimum value among the candidate feature similarities is used as the feature similarity of the candidate region.

[0128] Optionally, the second bare soil identification result determination module 340 is used to:

[0129] Determine a single bare soil area in the single characteristic area according to the first bare soil identification result;

[0130] Determining bare soil characteristic information according to bare soil color characteristics and bare soil brightness characteristics in the single bare soil area;

[0131] Dividing the mixed feature region into a plurality of candidate sub-regions;

[0132] If the sub-region color features and sub-region brightness features in the candidate sub-region in the mixed feature region successfully match the bare soil color features and bare soil brightness features, the candidate sub-region is determined to be a mixed bare soil region in the mixed feature region, and a second bare soil identification result is determined based on the mixed bare soil region.

[0133] Optionally, the first bare soil identification result determination module 330 is used to:

[0134] Determining an image feature vector of the single feature area;

[0135] Determine the region similarity of two single feature regions respectively according to the image feature vectors of the single feature regions;

[0136] If the region similarity is greater than a preset similarity threshold, it is determined that the region labels of the two single feature regions are the same;

[0137] Otherwise, determining that the region labels of the two single feature regions are different, until the region label of each single feature region is determined;

[0138] A single characteristic region with the same region label is input into the classifier to obtain a first bare soil recognition result corresponding to the region label.

[0139] Optionally, the device further comprises:

[0140] A boundary determination module, used for determining the boundary of the single feature area and the mixed feature area before determining the first bare soil identification result of the single feature area based on the pre-trained classifier;

[0141] An image feature determination module, used for extending along the boundary to the mixed feature area to determine the image features of the extended area;

[0142] A first processing module, configured to determine that the extended area is a single feature area if the similarity between the image feature of the extended area and the image feature of the single feature area adjacent to the boundary is greater than a preset threshold;

[0143] The second processing module is used for stopping the expansion otherwise; and updating the boundaries of the single feature area and the mixed feature area according to the expansion position;

[0144] An updated mixed feature region determination module is used to determine an updated mixed feature region according to the boundaries of the updated single feature region and the mixed feature region;

[0145] The final feature area determination module is used to take the updated mixed feature area as the image to be identified, continue to divide the image to be identified into multiple candidate areas, and determine the feature consistency judgment results of each candidate area according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is that there is no single feature area, then determine the final mixed feature area and the final single feature area.

[0146] Optionally, the device further comprises:

[0147] An image feature vector determination module, used for determining the image feature vector of the single feature region before extending along the boundary to the mixed feature region;

[0148] A region similarity determination module, used to respectively determine the region similarities of two adjacent single feature regions according to the image feature vectors of the single feature region;

[0149] A merged single feature region determination module is used to merge adjacent single feature regions according to the region similarity to obtain a merged single feature region;

[0150] Accordingly, the first processing module is used to:

[0151] If the similarity between the image features of the extended area and the image features of the merged single feature area adjacent to the boundary is greater than a preset threshold, the extended area is determined to be a single feature area.

[0152] A bare soil identification device provided in an embodiment of the present invention can execute a bare soil identification method provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0153] Embodiment 4

[0154] Figure 6A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0155] like Figure 6 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0156] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0157] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a bare soil identification method.

[0158] In some embodiments, the bare soil identification method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the bare soil identification method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the bare soil identification method in any other appropriate manner (e.g., by means of firmware).

[0159] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0160] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0161] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0162] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0163] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0164] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0165] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0166] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A bare soil identification method, characterized in that: The method comprises: Dividing the image to be identified in the area to be identified into a plurality of candidate areas; Determine the feature consistency judgment result of each candidate region according to the image features in the candidate region, and determine whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment result of the candidate region; Determining a first bare soil identification result of the single characteristic area based on a pre-trained classifier; Determining bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result; The bare soil identification result of the area to be identified is determined according to the first bare soil identification result and the second bare soil identification result.

2. The method according to claim 1, characterized in that: Determining the feature consistency judgment results of each candidate region according to the image features in the candidate region respectively includes: Dividing the candidate region into a plurality of sub-regions; Determine the image feature vector of each sub-region respectively; wherein the image feature vector includes at least a color dimension feature and / or a texture dimension feature; Determining the feature similarity of the candidate region according to the image feature vector of the sub-region; If the feature similarity is greater than a preset similarity threshold, determining that the feature consistency judgment result of the candidate region has feature consistency; Otherwise, determining that the feature consistency judgment result of the candidate region does not have feature consistency; Accordingly, determining whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment result of the candidate region includes: If the feature consistency judgment result of the candidate region is that the candidate region has feature consistency, the candidate region is determined to be a single feature region; If the feature consistency judgment result of the candidate region is that there is no feature consistency, the candidate region is determined to be a mixed feature region.

3. The method according to claim 2, characterized in that Determining the feature similarity of the candidate region according to the image feature vector of the sub-region includes: Inputting the image feature vector of the sub-region into a pre-trained consistency neural network discriminator, fusing and extracting the image feature vector through the network, and obtaining an intermediate feature vector corresponding to the image feature vector; Determine the candidate feature similarity between two intermediate feature vectors respectively; The minimum value among the candidate feature similarities is used as the feature similarity of the candidate region.

4. The method according to claim 1, characterized in that: Determining bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result includes: Determine a single bare soil area in the single characteristic area according to the first bare soil identification result; Determining bare soil characteristic information according to bare soil color characteristics and bare soil brightness characteristics in the single bare soil area; Dividing the mixed feature region into a plurality of candidate sub-regions; If the sub-region color features and sub-region brightness features in the candidate sub-region in the mixed feature region successfully match the bare soil color features and bare soil brightness features, the candidate sub-region is determined to be a mixed bare soil region in the mixed feature region, and a second bare soil identification result is determined based on the mixed bare soil region.

5. The method according to claim 1, characterized in that: Determining a first bare soil identification result of the single characteristic area based on a pre-trained classifier includes: Determining an image feature vector of the single feature area; Determine the region similarity of two single feature regions respectively according to the image feature vectors of the single feature regions; If the region similarity is greater than a preset similarity threshold, it is determined that the region labels of the two single feature regions are the same; Otherwise, determining that the region labels of the two single feature regions are different, until the region label of each single feature region is determined; A single characteristic region with the same region label is input into the classifier to obtain a first bare soil recognition result corresponding to the region label.

6. The method according to claim 1, characterized in that Before determining the first bare soil identification result of the single characteristic area based on the pre-trained classifier, the method further includes: Determine the boundaries of single feature areas and mixed feature areas; Extending toward the mixed feature region along the boundary to determine image features of the extended region; If the similarity between the image feature of the extended area and the image feature of the single feature area adjacent to the boundary is greater than a preset threshold, the extended area is determined to be a single feature area; Otherwise, stop expanding; and update the boundaries of the single feature area and the mixed feature area according to the expanded position; Determine an updated mixed feature region according to the boundaries of the updated single feature region and the mixed feature region; The updated mixed feature area is used as the image to be identified, and the image to be identified is continued to be divided into multiple candidate areas. The feature consistency judgment results of each candidate area are determined according to the image features in the candidate area, until the feature consistency judgment result of the candidate area is that there is no single feature area, then the final mixed feature area and the final single feature area are determined.

7. The method according to claim 6, characterized in that Before extending along the boundary toward the mixed feature region, the method further includes: Determining an image feature vector of the single feature area; Determine the region similarity of two adjacent single feature regions respectively according to the image feature vectors of the single feature region; Merging adjacent single feature regions according to the region similarity to obtain a merged single feature region; Correspondingly, if the similarity between the image feature of the extended area and the image feature of the single feature area adjacent to the boundary is greater than a preset threshold, determining that the extended area is a single feature area includes: If the similarity between the image features of the extended area and the image features of the merged single feature area adjacent to the boundary is greater than a preset threshold, the extended area is determined to be a single feature area.

8. A bare soil identification device, characterized in that: The device comprises: A candidate region division module is used to divide the image to be identified into a plurality of candidate regions; A feature consistency judgment module, used to determine the feature consistency judgment results of each candidate region according to the image features in the candidate region, and determine whether the candidate region is a single feature region or a mixed feature region according to the feature consistency judgment results of the candidate region; A first bare soil identification result determination module, used to determine a first bare soil identification result of the single feature area based on a pre-trained classifier; A second bare soil identification result determination module, configured to determine bare soil information in the mixed characteristic area according to the bare soil characteristic information as a second bare soil identification result; The bare soil identification result determination module is used to determine the bare soil identification result of the area to be identified according to the first bare soil identification result and the second bare soil identification result.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the bare soil identification method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the bare soil identification method described in any one of claims 1-7 when executed.

Images