A shunt method, shunt device, shunt system and processor for inspecting images

Abstract

Embodiments of the present application provide a kind of for the shunt method of inspection image, shunt device, shunt system and processor, belong to image recognition field.The shunt method for the inspection image includes: obtaining the multiple frames of inspection image of target inspection area that wearable image acquisition equipment is collected;Extract the feature point of inspection image;The feature point of adjacent frame of inspection image is matched, to obtain interframe matching feature point;According to interframe matching feature point, the interframe movement visual angle of wearable image acquisition equipment is determined;According to interframe movement visual angle and interframe matching feature point, determine the effective inspection image in multiple frames of inspection image;Effective inspection image is pushed to server.Adopt the scheme of embodiment of the present application can improve real-time performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image recognition, in particular to a shunting method, a shunting device, a shunting system and a processor for image inspection. BACKGROUND

[0002] It has become increasingly popular to perform equipment maintenance inspection through wearable image collection devices (for example, a helmet with a camera). However, only about one-third of the time of the maintenance inspection process is actually valuable inspection, and the remaining time is ineffective. If a user pushes the entire inspection process video to the server, the user needs to occupy a large amount of mobile data traffic, and the long-term accumulation of traffic fees is also a considerable expense. Therefore, the method commonly used in the prior art is to not push during the inspection process, and to push in a WiFi environment. However, this method has the problem of low real-time performance. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a shunting method, a processor, a shunting device, a shunting system and a storage medium for image inspection, to solve the problem of low real-time performance in the prior art.

[0004] To achieve the above-mentioned purpose, the first aspect of the embodiments of the present application provides a shunting method for image inspection, the shunting method comprising:

[0005] obtaining a plurality of inspection images of a target inspection area collected by a wearable image collection device;

[0006] extracting feature points of the inspection images;

[0007] matching the feature points of the inspection images of adjacent frames to obtain inter-frame matching feature points;

[0008] determining an inter-frame movement perspective of the wearable image collection device according to the inter-frame matching feature points;

[0009] determining effective inspection images in the plurality of inspection images according to the inter-frame movement perspective and the inter-frame matching feature points;

[0010] pushing the effective inspection images to a server.

[0011] In the embodiments of the present application, the feature points of the inspection images are extracted, and the method further comprises: extracting a feature vector corresponding to the feature points; and matching the feature points of the inspection images of adjacent frames to obtain inter-frame matching feature points, which comprises: determining a similarity of the feature vectors; and matching the feature points of the inspection images of adjacent frames according to the similarity to obtain the inter-frame matching feature points.

[0012] In the embodiment of the present application, the effective inspection image in the multiple inspection images is determined according to the inter-frame moving view angle and the inter-frame matching feature points, comprising: based on a preset classification algorithm, the inspection images are classified according to the inter-frame moving view angle and the inter-frame matching feature points to obtain a classification result; and the effective inspection image is determined according to the classification result.

[0013] In the embodiment of the present application, the effective inspection image is determined according to the classification result, comprising: based on a preset inspection target recognition algorithm, whether the inspection image includes an inspection target is recognized according to the feature points to obtain a recognition result; an image quality index value of the inspection image is determined, wherein the image quality index value includes at least one of a gray mean value, a gray standard deviation, a gray average gradient and a gray entropy; and the effective inspection image is determined according to the recognition result and / or the image quality index value, the classification result.

[0014] In the embodiment of the present application, the effective inspection image is determined according to the recognition result and / or the image quality index value, the classification result, comprising: the recognition result and / or the image quality index value, the classification result are weighted to obtain a weighted result; and the inspection image is determined as the effective inspection image in the case that the weighted result reaches a preset threshold.

[0015] In the embodiment of the present application, based on the preset classification algorithm, the inspection images are classified according to the inter-frame moving view angle and the inter-frame matching feature points to obtain the classification result, comprising: a feature vector corresponding to the inter-frame matching feature points is extracted; the feature vector corresponding to the inter-frame moving view angle and the inter-frame matching feature points in a preset time period is taken as a sequence feature; and based on the preset classification algorithm, the inspection images are classified according to the sequence feature to obtain the classification result.

[0016] In the embodiment of the present application, the classification result includes: static, effective moving and invalid moving.

[0017] In the embodiment of the present application, the similarity includes distance similarity and / or cosine similarity.

[0018] The second aspect of the embodiment of the present application provides a processor configured to execute the shunting method for inspection images according to the above.

[0019] The third aspect of the embodiment of the present application provides a shunting device for inspection images, comprising: a wearable image acquisition device; and a processor according to the above.

[0020] The fourth aspect of the embodiment of the present application provides a shunting system for inspection images, comprising: a processor according to the above; and a server in communication with the processor, configured to receive the effective inspection image pushed by the processor.

[0021] The fifth aspect of the embodiment of the present application provides a machine readable storage medium, which stores instructions, and the instructions make the processor execute the method for distributing the inspection images according to the above description when executed by the processor.

[0022] The method for distributing the inspection images, by acquiring the multiple frames of inspection images of the target inspection area collected by the wearable image collection device, extracting the feature points of the inspection images, matching the feature points of the adjacent frames of inspection images to obtain the inter-frame matching feature points, determining the inter-frame moving visual angle of the wearable image collection device according to the inter-frame matching feature points, determining the effective inspection image in the multiple frames of inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points, and pushing the effective inspection image to the server. The above scheme matches the feature points of the adjacent frames of inspection images to obtain the inter-frame matching feature points, determines the inter-frame moving visual angle according to the inter-frame matching feature points, and identifies the effective inspection image in the multiple frames of inspection images according to the inter-frame moving visual angle. The inter-frame moving visual angle is considered and added to the identification and judgment process of the effective inspection image, which can improve the identification accuracy of the effective inspection image. All the inspection images do not need to be pushed to the server, which reduces the traffic cost of the user. The storage capacity of the server end is reduced, the real-time of the pushing is improved, the user can realize the pushing without manual operation, and the safety of the user inspection process is further improved.

[0023] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings are included to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the embodiments of the present application together with the following specific implementation, but do not constitute a limitation of the embodiments of the present application. In the drawings:

[0025] Figure 1 The flowchart of the method for distributing the inspection images in the embodiment of the present application is schematically shown;

[0026] Figure 2 The flowchart of the method for distributing the inspection images in another embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0027] The specific implementation of the embodiments of the present application is described in detail below in combination with the drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiments of the present application, and is not used to limit the embodiments of the present application.

[0028] The flow control method commonly used in the prior art includes the following two kinds: one is not to push the flow in the inspection process, and to store in the WiFi environment and then push the flow, using this method, the video is not pushed to the server in real time, and there is a problem of low real-time performance; two is that the user manually controls whether to push the flow in the inspection process, which increases the workload of the user, and in some scenarios (such as high altitude), the user cannot operate due to safety considerations.

[0029] Figure 1 The flowchart of the flow control method for the inspection image in the embodiment of the application is schematically shown. As shown in the figure, Figure 1 In the embodiment of the application, a flow control method for inspection image is provided, which is taken as an example of a processor, and the flow control method can include the following steps:

[0030] Step S102, acquiring a plurality of inspection images of a target inspection area collected by a wearable image collection device.

[0031] It can be understood that the target inspection area is an object or area that needs to be inspected. The wearable image collection device is an image collection device (for example, a helmet with a camera) installed on a wearable device. The inspection image is an image about the target inspection area in the inspection process.

[0032] Specifically, the processor can acquire a plurality of inspection images of the target inspection area collected by the wearable image collection device (for example, a helmet with a camera).

[0033] Step S104, extracting feature points of the inspection image.

[0034] Specifically, the processor can extract feature points through a feature detection algorithm, which can include but is not limited to sift, surf, fast, orb algorithm, etc. Taking the fast feature detection algorithm as an example, the fast feature detection algorithm can extract the feature points of the inspection image according to the pixel point gradient.

[0035] In addition, a pre-trained feature point extraction model can also be used to extract the feature points of each inspection image in the embodiment.

[0036] Step S106, matching the feature points of the adjacent frame inspection images to obtain inter-frame matching feature points.

[0037] It can be understood that the inter-frame matching feature points are feature points on the adjacent frame inspection images with a similarity reaching a certain set threshold.

[0038] Specifically, the processor can match the feature points of the two continuous inspection images based on the similarity, so as to determine the inter-frame matching feature points on the adjacent frame inspection images.

[0039] In one embodiment, the feature points of the inspection images are extracted, and the method further comprises: extracting feature vectors corresponding to the feature points; and matching the feature points of the inspection images of the adjacent frames to obtain inter-frame matching feature points, including: determining the similarity of the feature vectors; and matching the feature points of the inspection images of the adjacent frames according to the similarity to obtain the inter-frame matching feature points.

[0040] It can be understood that the feature vectors include relevant parameter features corresponding to the feature points, such as brightness information and / or color information.

[0041] Specifically, the processor can first extract the feature vectors corresponding to the feature points, determine the similarity of the feature vectors, for example, distance similarity, and match the feature points of the inspection images of the adjacent frames according to the similarity of the feature vectors to obtain the inter-frame matching feature points, for example, the feature points whose distance similarity reaches a preset distance similarity threshold are the inter-frame matching feature points.

[0042] In step S108, the inter-frame movement visual angle of the wearable image acquisition device is determined according to the inter-frame matching feature points.

[0043] It can be understood that the inter-frame movement visual angle is the movement angle of the visual angle of the wearable image acquisition device when collecting the inspection images of the adjacent frames, that is, the spatial movement visual angle of the inspection images of the adjacent frames.

[0044] Specifically, the processor can calculate the spatial movement visual angle of the adjacent frames, that is, the inter-frame movement visual angle of the wearable image acquisition device, according to the spatial positional relationship of the inter-frame matching feature points, which can be determined by a corresponding model or algorithm, and details are not described here.

[0045] In step S110, the effective inspection image in the multiple inspection images is determined according to the inter-frame movement visual angle and the inter-frame matching feature points.

[0046] It can be understood that the effective inspection image is an inspection image that includes an inspection target (for example, a point) and the corresponding inter-frame movement visual angle is within a preset visual angle range, which has actual inspection value or effect. The preset visual angle range is a pre-set appropriate movement visual angle range, that is, the inter-frame movement visual angle is not too large or too small. Understandably, when the inter-frame movement visual angle is too large, it means that the wearable image acquisition device is in a fast motion state, and when the inter-frame movement visual angle is too small, it means that the wearable image acquisition device is in a static state. If a certain part is being inspected, the wearable image acquisition device is most likely in slow motion, so the inter-frame movement visual angle is not too large or too small, so the movement visual angle itself is a key feature that can distinguish the effective inspection image.

[0047] Specifically, the processor can determine whether the inspection image is an effective inspection image based on the pre-trained effective inspection image recognition model, taking the inter-frame moving visual angle and the inter-frame matching feature points as inputs of the effective inspection image recognition model, and obtaining an output result of the effective inspection image recognition model.

[0048] In step S112, the effective inspection image is pushed to the server.

[0049] Specifically, after determining the effective inspection image, the processor can push the effective inspection image in the multiple inspection images to the server.

[0050] The above-mentioned method for splitting the inspection image can obtain multiple inspection images of a target inspection area collected by a wearable image collection device, extract feature points of the inspection images, match the feature points of adjacent inspection images to obtain inter-frame matching feature points, determine the inter-frame moving visual angle of the wearable image collection device according to the inter-frame matching feature points, determine the effective inspection image in the multiple inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points, and push the effective inspection image to the server. The above-mentioned scheme matches the feature points of adjacent inspection images to obtain inter-frame matching feature points, determines the inter-frame moving visual angle according to the inter-frame matching feature points, and identifies the effective inspection image in the multiple inspection images according to the inter-frame moving visual angle. The inter-frame moving visual angle is considered and added to the identification and judgment process of the effective inspection image, which can improve the identification accuracy of the effective inspection image. All inspection images do not need to be pushed to the server, which reduces the flow cost of the user. The effective inspection image with real value can be pushed to the server, which reduces the storage capacity of the server, improves the real-time performance of the pushing, and further improves the safety of the user inspection process without manual operation of the user.

[0051] In one embodiment, determining the effective inspection image in the multiple inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points includes: classifying the inspection image according to the inter-frame moving visual angle and the inter-frame matching feature points based on a preset classification algorithm to obtain a classification result; and determining the effective inspection image according to the classification result.

[0052] It can be understood that the classification algorithm is a pre-determined algorithm for classifying the inspection image. Specifically, the classification algorithm can include but is not limited to classification svm, decision tree, perception machine, etc. The classification result can include effective inspection image and ineffective inspection image, for example.

[0053] Specifically, the processor can classify the inspection images according to the inter-frame moving angle and the inter-frame matching feature points based on a preset classification algorithm to obtain a classification result (i.e., a motion classification result), wherein the classification result can include valid inspection images and invalid inspection images, and then the valid inspection images in the classification result can be determined as the final valid inspection images.

[0054] In one embodiment, the classification result can include: still, valid movement, and invalid movement.

[0055] It can be understood that still means a still inspection image, that is, an inspection image with an inter-frame moving angle of 0, valid movement means a valid inspection image, that is, an inspection image with an inter-frame moving angle that is not too large nor equal to 0, and invalid movement means an invalid inspection image, that is, an inspection image with a large inter-frame moving angle. In some embodiments, the still inspection images and the valid inspection images in the above classification result can be determined whether to be uniformly classified as valid inspection images by determining the distance in time. The still inspection images that are close in time (e.g., less than a certain set time distance threshold) can be classified as valid inspection images, otherwise the still inspection images are not classified as valid inspection images.

[0056] In one embodiment, determining the valid inspection images according to the classification result includes: determining whether the inspection images include the inspection target according to the feature points based on a preset inspection target recognition algorithm to obtain a recognition result; determining an image quality indicator value of the inspection image, wherein the image quality indicator value includes at least one of a gray mean value, a gray standard deviation, a gray average gradient, and a gray entropy; and determining the valid inspection images according to the recognition result and / or the image quality indicator value, the classification result.

[0057] It can be understood that the inspection target recognition algorithm is used to identify whether the inspection images include the inspection target, which can be determined in advance. Specifically, the inspection target recognition algorithm can include but is not limited to target recognition ssd, yolo, RetinaNet, centernet, etc.

[0058] Specifically, the processor can identify whether the inspection target is included in the inspection image according to the feature points of the inspection image by a preset inspection target recognition algorithm to obtain an identification result, and then the processor can determine an image quality index value (including at least one of a gray mean value, a gray standard deviation, a gray average gradient, and a gray entropy) of the inspection image, and then determine the effective inspection image according to the identification result and / or the image quality index value and the classification result, that is, the effective inspection image can be determined according to at least one of the identification result and the image quality index value and the classification result, that is, the effective inspection image can be determined according to the identification result and the classification result, for example, only when the identification result is that the inspection image includes the inspection target and the classification result is that the effective inspection image is determined as the effective inspection image; the effective inspection image can also be determined according to the image quality index value and the classification result, for example, only when the image quality index value reaches a preset index value and the classification result is that the effective inspection image is determined as the effective inspection image; the effective inspection image can also be determined according to the identification result, the image quality index value, and the classification result, for example, only when the identification result is that the inspection image includes the inspection target, the image quality index value reaches the preset index value, and the classification result is that the effective inspection image is determined as the effective inspection image.

[0059] Further, regarding the image quality index value, taking the gray standard deviation as an example, when the gray standard deviation is too small, it indicates that the image is too blurred, and therefore only when the gray standard deviation is within a suitable range can it indicate that the image quality is good, and the same applies to other index values.

[0060] In the embodiment of the application, the effective inspection image is determined according to at least one of the identification result of the inspection target and the image quality index value and the motion classification result, which can further improve the judgment accuracy of the effective inspection image, reduce the push error rate, and achieve accurate push.

[0061] In one embodiment, the effective inspection image is determined according to the identification result and / or the image quality index value and the classification result, including: weighting the identification result and / or the image quality index value and the classification result to obtain a weighted result; and in the case that the weighted result reaches a preset threshold, determining the inspection image as an effective inspection image.

[0062] It can be understood that the preset threshold is a pre-set weighted result threshold, and when it is greater than the preset threshold, the inspection image can be determined as an effective inspection image.

[0063] Specifically, the processor can weight at least one of the identification result of whether the inspection target is included in the inspection image and the image quality index value and the classification result, so as to obtain a weighted result, that is, a comprehensive score result, and in the case that the weighted result reaches a preset threshold, the inspection image can be determined as an effective inspection image, so as to be pushed to the server, otherwise, no push is performed.

[0064] In one embodiment, based on a preset classification algorithm, the inspection images are classified according to the inter-frame moving view angle and the inter-frame matching feature points to obtain a classification result, including: extracting the feature vectors corresponding to the inter-frame matching feature points; taking the feature vectors corresponding to the inter-frame moving view angle and the inter-frame matching feature points in a preset time period as sequence features; and based on a preset classification algorithm, classifying the inspection images according to the sequence features to obtain a classification result.

[0065] It can be understood that the preset time period is a pre-set fixed time period, i.e. a fixed time window. The inspection judgment is not a state judgment, but a process judgment, which depends on a period of video input, rather than the last frame, so the sequence features in the preset time period need to be obtained, which can include the inter-frame moving view angle and the feature vectors and other parameters.

[0066] Specifically, the feature points and the feature vectors have a one-to-one correspondence, and after determining the inter-frame matching feature points, the processor can extract the feature vectors corresponding to the inter-frame matching feature points, and can take the feature vectors corresponding to the inter-frame moving view angle and the inter-frame matching feature points in the preset time period as sequence features, and take the sequence features as the input of the classification algorithm, so that the inspection images in the preset time period can be classified according to the sequence features to obtain the corresponding classification result.

[0067] In one embodiment, the similarity can include distance similarity and / or cosine similarity.

[0068] It can be understood that the processor can match the feature points of the inspection images of adjacent frames according to the distance similarity to obtain the inter-frame matching feature points, further, for example, the feature points of the adjacent frame inspection images with the distance similarity reaching a preset distance similarity threshold are determined as the inter-frame matching feature points; or the feature points of the adjacent frame inspection images can be matched according to the cosine similarity to obtain the inter-frame matching feature points, further, for example, the feature points of the adjacent frame inspection images with the cosine similarity reaching a preset cosine similarity threshold are determined as the inter-frame matching feature points; or the feature points of the adjacent frame inspection images can be matched according to the distance similarity and the cosine similarity to obtain the inter-frame matching feature points, further, for example, the feature points of the adjacent frame inspection images with the distance similarity reaching a preset distance similarity threshold and the cosine similarity reaching a preset cosine similarity threshold are determined as the inter-frame matching feature points.

[0069] In one specific embodiment, as shown in Figure 2 a method for inspection image shunting is provided, including:

[0070] 1. Feature point extraction: extracting feature points and feature vectors of video frames according to brightness, color, shape and other information.

[0071] 2. Inter-frame matching feature points: match feature points of two consecutive frames according to the similarity of feature vectors, the similarity measurement method includes but is not limited to distance or cosine similarity.

[0072] 3. Inter-frame view movement calculation: calculate the spatial view movement of adjacent frames from the spatial position relationship of matching feature points.

[0073] 4. Classification feature selection: select a fixed time window, and take the inter-frame view movement angle and the feature vector of the inter-frame matching feature point as the sequence feature.

[0074] 5. Motion classification: the features in step 4 are taken as input to train a sequence classification algorithm, which can be divided into three categories: static, valid movement and invalid movement.

[0075] 6. Inspection target: use the feature vector in step 1 as input to obtain whether each frame contains an inspection target through a classification or detection algorithm.

[0076] 7. Image quality evaluation: calculate quality evaluation index values such as the average gray value of the video frame, the standard deviation of the gray value, the average gradient of the gray value, and the entropy of the gray value.

[0077] 8. Comprehensive judgment: select the results of steps 5, 6 and 7 as input to determine whether it is a valid inspection through an artificial rule comprehensive scoring method or a classification algorithm.

[0078] 9. For the video frame determined as valid inspection, push it to the server, and do not push the frames classified as static and invalid movement, thereby realizing flow control, saving server storage and reducing battery consumption.

[0079] In the technical scheme of the embodiment of the application, the key points (i.e., feature points) of each frame are first extracted, then the key points (i.e., feature points) are used to estimate the movement amplitude of the wearable image acquisition device (for example, a helmet with a camera), and then the classification judgment of valuable inspection images (or video frames) is made according to the feature vectors corresponding to the key points (i.e., feature points). The above technical scheme calculates the movement view based on feature point matching, and then determines the valid inspection image based on motion classification, whether the target is inspected and image quality indicators to realize intelligent flow control, greatly reducing the cost of traffic fees and the storage cost of the server side. The whole process does not require the user to perform other additional operations, is safe and convenient, and can realize real-time video streaming without long delay.

[0080] The embodiment of the application provides a processor configured to perform the shunt method for inspecting images in the above embodiments.

[0081] The embodiment of the present application provides a shunting device for patrolling images, comprising: a wearable image collection device; and a processor according to the above-mentioned embodiment.

[0082] The embodiment of the present application provides a shunting system for patrolling images, comprising: a processor according to the above-mentioned embodiment; and a server, in communication with the processor, for receiving the effective patrolling image pushed by the processor.

[0083] The embodiment of the present application provides a machine readable storage medium, which stores instructions, and the instructions cause a processor to execute the shunting method for patrolling images according to the above-mentioned embodiment when the instructions are executed by the processor.

[0084] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product in the form of one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0085] The present application is described with reference to flowcharts and / or block diagrams according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks

[0086] These computer program instructions can also be stored in a computer readable storage medium, which can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks

[0087] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1

[0088] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0089] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) about which the processor can execute instructions. The memory can also include non-volatile memory, such as read only memory (ROM), electrically programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), erasable programmable read only memory (EPROM), flash memory, or a combination of non-volatile memories in different types. The memory can also include a compact disk read only memory (CD-ROM), digital versatile disk (DVD), Blu-ray, or another non-transitory computer readable medium, which is non-volatile and non-transitory in nature, but volatile in that it can lose its content if the power to the computer is turned off or if the computer crashes. The memory is an example of a computer readable medium.

[0090] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically programmable read only memory (EEPROM), flash memory or other memory technology, compact disk read only memory (CD-ROM), digital versatile disk (DVD), or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media such as modulated data signals and carriers.

[0091] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0092] ​​The above merely provides an example of the present application, and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall fall into the scope of claims of the present application.

Claims

1. A shunting method for inspecting images, characterized in that, The shunting method comprises: acquiring a plurality of inspection images of a target inspection area collected by a wearable image collection device; extracting feature points of the inspection images; matching the feature points of the inspection images of adjacent frames to obtain inter-frame matching feature points; determining an inter-frame moving visual angle of the wearable image collection device according to the inter-frame matching feature points; determining an effective inspection image in the plurality of inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points; pushing the effective inspection image to a server; wherein the determining the effective inspection image in the plurality of inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points comprises: classifying the inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points based on a preset classification algorithm to obtain a classification result; and determining the effective inspection image according to the classification result; the classifying the inspection images according to the inter-frame moving visual angle and the inter-frame matching feature points based on the preset classification algorithm to obtain the classification result comprises: extracting feature vectors corresponding to the inter-frame matching feature points; taking the feature vectors corresponding to the inter-frame moving visual angle and the inter-frame matching feature points in a preset time period as sequence features; and classifying the inspection images according to the sequence features based on the preset classification algorithm to obtain the classification result; the classification result comprises: stillness, effective movement, and ineffective movement.

2. The split-flow method of claim 1, wherein, The extracting the feature points of the inspection images further comprises: extracting feature vectors corresponding to the feature points; the matching the feature points of the inspection images of adjacent frames to obtain inter-frame matching feature points comprises: determining a similarity of the feature vectors; matching the feature points of the inspection images of adjacent frames according to the similarity to obtain inter-frame matching feature points.

3. The split-flow method of claim 1, wherein, the determining the effective inspection image according to the classification result comprises: identifying whether the inspection images include an inspection target based on the feature points according to a preset inspection target identification algorithm to obtain an identification result; determining an image quality index value of the inspection images, wherein the image quality index value comprises at least one of a gray mean value, a gray standard deviation, a gray average gradient, and a gray entropy; determining the effective inspection image according to the identification result and / or the image quality index value, and the classification result.

4. The split-flow method of claim 3, wherein, the determining the effective inspection image according to the identification result and / or the image quality index value, and the classification result comprises: weighting the identification result and / or the image quality index value, and the classification result to obtain a weighting result; in a case where the weighting result reaches a preset threshold, determining the inspection image as an effective inspection image.

5. The split-flow method of claim 2, wherein, the similarity comprises a distance similarity and / or a cosine similarity.

6. A processor, comprising: A device configured to perform the shunting method for inspection images according to any one of claims 1 to 5.

7. A shunt device for inspecting an image, characterized by comprises: a wearable image collection device; and a processor according to claim 6.

8. A shunting system for inspecting images, characterized in that comprises: a processor according to claim 6; and A server, in communication with the processor, for receiving the effective inspection image pushed by the processor.

9. A machine-readable storage medium having stored thereon instructions, the instructions being executable by a machine to cause the machine to perform operations comprising: The instruction, when executed by a processor, causes the processor to perform the method for splitting an inspection image according to any one of claims 1 to 5.

Images