Highway construction safety risk intelligent identification platform based on artificial intelligence

Abstract

The invention discloses a highway construction safety risk intelligent identification platform based on artificial intelligence, and the platform comprises a terminal which is used for receiving risk identification information and then carrying out the information presentation; the server is used for receiving the risk identification information and analyzing, storing and transmitting the risk identification information; the AI analysis box is used for receiving the image information and analyzing and processing the image information, and the process of analyzing and processing the image comprises the following steps: receiving a plurality of images in a first preset time period, randomly selecting a plurality of areas in one image, correspondingly endowing other images with the areas to be selected, and storing the selected areas in the first preset time period; performing feature recognition of the type on the images in the to-be-selected area of each image, counting the number of pre-selected areas of the plurality of images according to area positions, and defining the area position with the maximum number of pre-selected areas and exceeding a first preset proportion as a feature selection area of the type; and the camera is used for collecting image information. Compared with the prior art, the method can achieve the precise recognition of the safety risk.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and particularly relates to an intelligent identification platform for highway construction safety risks based on artificial intelligence. Background Art

[0002] The current development of computer vision has led to the interaction and integration of multiple disciplines. It mainly studies how a computer can provide rich information through digital images or videos to support and achieve high-level understanding. Due to the potential of computer vision in the automation and sustainable monitoring of the engineering field, it has attracted an increasing amount of research interest in the engineering industry. Currently, the development of deep learning is relatively fast, the amount of image data has increased massively, and the computer processing power has also been improved at a high speed. Computer vision can use deep learning-based technologies to establish models, process and analyze images obtained by different means in a timely manner, and provide rich information about the construction site to provide an accurate and comprehensive understanding of construction activities. The essence of deep learning is actually to build a deep network model with many hidden layers. After being trained with a large amount of data, it can learn effective target features, and accordingly, the effect of classifying or predicting other new targets can be improved. Therefore, applying the convolutional neural network in deep learning to the field of target detection can greatly improve the target detection ability compared with traditional manual supervision, and can perform automatic monitoring and detection on the construction site, providing technical support for the intelligent management of the construction site.

[0003] There are numerous construction activities and tasks at the construction site, a large number of construction entity targets, and complex construction scenarios. Traditional supervision of the construction site is often carried out through manual supervision, which results in the information faced by the construction site management personnel being cumbersome and complex, with a large amount of data and being difficult to process efficiently.

[0004] In view of this, a special intelligent identification platform for highway construction safety risks based on artificial intelligence is provided. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent identification platform for highway construction safety risks based on artificial intelligence, which can achieve accurate identification of safety risks.

[0006] The above technical purpose of the present invention is achieved through the following technical solutions: An intelligent identification platform for highway construction safety risks based on artificial intelligence, comprising: A terminal, which is used for presenting information after receiving the risk identification information sent by the server; A server, which is used for receiving the risk identification information of the AI analysis box, and analyzing, storing, and transmitting the risk identification information; The AI analysis box is used to receive the image information collected by the camera and analyze and process the image information. The process of analyzing and processing the image includes: receiving multiple images within a first preset time period, randomly selecting multiple regions in one of the images, defining them as candidate regions, and assigning the candidate regions to other images correspondingly. For each image, perform feature recognition of this type on the image within the candidate region, define the candidate region where the feature of this type is recognized as a preselected region, count the number of preselected regions of multiple images according to the region position, and define the region position with the largest number of preselected regions and exceeding the first preset ratio as the selected region of this type of feature. The first preset ratio refers to the ratio of the number of preselected regions in a certain region position preset in advance to the total number of images. The camera is used to be installed at the construction site to collect image information.

[0007] In a preferred embodiment, the terminal includes a PC terminal, a mobile phone terminal, and a large screen terminal.

[0008] In a preferred embodiment, there are multiple AI analysis boxes and multiple cameras.

[0009] In a preferred embodiment, the first preset time period is set to 0.5 - 1 s, and the first preset ratio is set to 0.8 - 0.9.

[0010] In a preferred embodiment, the feature of this type is the feature of wearing a safety helmet, the feature of wearing a safety belt, or the feature of taking a rest on the construction platform.

[0011] In a preferred embodiment, the AI analysis box is also used to regularly clear the information of the selected region of this type of feature and then re - select it.

[0012] In a preferred embodiment, the AI analysis box is also used to receive the image information collected by the camera and perform feature recognition on the people in the selected region of this type of feature in the image. If the feature of this type cannot be recognized, an alarm signal is sent to the terminal.

[0013] In a preferred embodiment, the AI analysis box is also used to re - select the selected region of this type of feature when the selection of the selected region of this type of feature fails.

[0014] In a preferred embodiment, the AI analysis box is further configured to, when the selection of the selected area of the current type of feature fails again, determine whether there are more than two selected areas of other feature types. If so, randomly select two selected areas of other feature types, receive multiple images within a second preset time period collected by the camera, identify the features of the current type within the two selected areas of other feature types selected in each image, and count the number of selected areas of other feature types in which the features of the current type are identified. If, in the two selected areas of other feature types selected, the number of images in which the features of the current type are identified in each selected area of other feature types reaches the second preset ratio, then define these two selected areas of other feature types as the jointly selected areas of the features of the current type.

[0015] In a preferred embodiment, the AI analysis box is further configured to, when the selection of the jointly selected area of the features of the current type fails, re-select the jointly selected area of the features of the current type.

[0016] In a preferred embodiment, the second preset time period is set to 0.5 - 1 s, and the second preset ratio is set to 0.6.

[0017] In a preferred embodiment, the AI analysis box is further configured to receive the image information collected by the camera and perform feature recognition on the people in the jointly selected area of the features of the current type in the image. If the features of the current type cannot be recognized in both areas of the jointly selected area of the features of the current type, an alarm signal is sent to the terminal.

[0018] Compared with the prior art, the present invention is equivalent to pre-selecting a specific area for feature recognition of a specific type within the field of view of the camera. When performing feature recognition of a specific type, it is no longer necessary to perform recognition in the entire image frame, thereby effectively reducing the recognition difficulty and the amount of calculation; the selected area of the features of the current type selected by this solution is the area that can best recognize the corresponding features, so it has the best recognition effect, thereby ensuring the accuracy of feature recognition. Compared with the previous method of performing recognition in the entire image, the accuracy has been effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic diagram of the architecture of an intelligent highway construction safety risk intelligent recognition platform based on artificial intelligence according to the present invention.

[0020] Figure 2 is a schematic diagram of an image for feature recognition during the process of an intelligent highway construction safety risk intelligent recognition platform based on artificial intelligence performing tasks according to the present invention.

[0021] Figure 3It is a schematic diagram of an image for joint feature recognition during the execution of tasks by an intelligent highway construction safety risk intelligent recognition platform based on artificial intelligence. In the figure

[0022] PC terminal 1; large screen terminal 2; mobile terminal 3; server 4; AI analysis box 5; camera 6. Specific implementation manners

[0023] The present invention will be further described in detail below with reference to the accompanying drawings.

[0024] This specific embodiment is only an interpretation of the present invention, and it does not limit the present invention. After reading this specification, those skilled in the art can make modifications without creative contributions to this embodiment as needed, but as long as it is within the scope of the claims of the present invention, it is protected by the patent law.

[0025] As Figures 1 to 3 shown, an intelligent highway construction safety risk intelligent recognition platform based on artificial intelligence includes: A terminal, configured to present information after receiving the risk recognition information sent by the server 4; The server 4, configured to receive the risk recognition information of the AI analysis box 5, and analyze, store, and transmit the risk recognition information; The AI analysis box 5, configured to receive the image information collected by the camera 6, and perform analysis and processing on the image information. The process of performing analysis and processing on the image includes: receiving multiple images within a first preset time period, randomly selecting multiple regions in one of the images, defining them as candidate regions, and assigning the candidate regions to other images. Perform this type of feature recognition on the images within the candidate regions of each image, define the candidate regions where this type of feature is recognized as preselected regions, count the number of preselected regions of multiple images according to the region positions, and define the region position with the largest number of preselected regions and exceeding the first preset ratio as the selected region of this type of feature. The first preset ratio refers to the ratio of the number of preselected regions at a certain region position preset in advance to the total number of images; The camera 6, configured to be installed at the construction site to collect image information.

[0026] An intelligent recognition platform for highway construction safety risks based on artificial intelligence in this embodiment, when performing tasks, is equivalent to pre-selecting a specific area for specific type of feature recognition within the field of view of the camera 6. When performing feature recognition of a specific type, it is no longer necessary to perform recognition in the entire image frame, thereby effectively reducing the recognition difficulty and computational amount; the selected area of this type of feature selected by this solution is the area that can best recognize the corresponding feature, thus having the best recognition effect, and being able to ensure the accuracy of feature recognition. Compared with the previous method of performing recognition in the entire image, the accuracy has been effectively improved.

[0027] Specifically, referring to Figure 1 , the circles represent the areas to be selected, the circles with slashes represent the pre-selected areas, and the black circles represent the selected areas of this type of feature. Among them, the pre-selected areas are blank circles before being defined as pre-selected areas, and the black circles are circles with slashes before being defined as the selected areas of this type of feature. Therefore, in the initial state, all the circles are areas to be selected, and then the pre-selected areas are gradually selected from these areas to be selected, and finally the selected areas of this type of feature are selected. The selected areas of this type of feature are the areas for recognizing the set type of feature during the actual recognition process.

[0028] Taking safety helmets as an example of this type of feature, receive multiple images within the first preset time period. The multiple images are specifically set to 10. In one of them, randomly select multiple areas and define them as areas to be selected. We randomly select 12 areas, so there are 12 areas to be selected. Assign the positions of these 12 areas to the other 9 images, so that each image has 12 areas to be selected with corresponding positions. Among these areas to be selected, perform feature recognition of this type, that is, perform feature recognition of safety helmet wearing. If the area is recognized as having the feature of safety helmet wearing, then this area is defined as a pre-selected area. After the recognition is completed, add up the number of pre-selected areas recognized at each position of all the images respectively, and then the number of pre-selected areas recognized at each position can be obtained. Select the position area with the largest number of pre-selected areas as the object to be selected for the possible selected area of this type of feature. As long as the condition is met that it exceeds the first preset ratio, that is to say, to be defined as the selected area of this type of feature among these 12 areas, it is required that more than the first preset ratio of the 10 images are defined as pre-selected areas at this area position, then this area position is determined as the selected area of this type of feature.

[0029] The above execution process is the formation process of the selected area of one type of feature. In actual applications, through the above steps, the selection of selected areas of multiple types can be achieved, thereby realizing the recognition of multiple features.

[0030] Specifically, the present type of feature is a safety helmet wearing feature, a safety belt wearing feature, or a construction platform rest feature. In practical applications, the types of features do not only include the above feature types, but can include multiple types of features.

[0031] In this embodiment, the terminal includes a PC terminal 1, a mobile phone terminal 3, and a large screen terminal 2, to achieve multi-terminal presentation of the risk identification results.

[0032] To achieve risk monitoring of multiple regions, multiple AI analysis boxes 5 and cameras 6 are provided.

[0033] Further, the first preset time period is set to 0.5 - 1 s. During this period, the reasonable number of images is 8 - 16. The first preset ratio is set to 0.8 - 0.9. At this ratio setting, it indicates a relatively high recognition accuracy, and thus can be defined as a qualified selected area of this type of special feature.

[0034] To update according to the on-site situation and avoid the invalidation of the selected area of this type of feature, the AI analysis box 5 is also used to regularly clear the information of the selected area of this type of feature and then re-select it.

[0035] The AI analysis box 5 is also used to receive the image information collected by the camera 6, and perform feature recognition on the people in the selected area of this type of feature in the image. If the feature of this type cannot be recognized, an alarm signal is sent to the terminal, that is, a normal risk identification process is realized.

[0036] Further, the AI analysis box 5 is also used to re-select the selected area of this type of feature when the selection of the selected area of this type of feature fails (the main situation is that the feature of this type is not recognized at any position, or does not meet the requirements of the first preset ratio).

[0037] Further, when the re-selection of the selected area of this type of feature fails, that is, it means that the feature recognition of this type cannot be realized at a certain area position, the AI analysis box 5 is also used to determine whether there are two or more other feature type selected areas. If so, two other feature type selected areas are randomly selected, and multiple images within the second preset time period collected by the camera 6 are received. The feature of this type within the two selected other feature type selected areas in each image is recognized, and the number of other feature type selected areas in which the feature of this type is recognized is counted. If for each of the two selected other feature type selected areas, the number of images in which the feature of this type is recognized reaches the second preset ratio, then these two other feature type selected areas are defined as the combined selected area of this type of feature.

[0038] The setting of the above process is to make a judgment through selected areas of other types when it is impossible to obtain a selected area of this type of feature that meets the requirements in the image. The possible reason may be that the position of the camera 6 results in a low recognition rate of this type of feature. Therefore, in this case, we use selected areas of other types of features to assist in the judgment. We select two selected areas of other feature types. When the number of images of this type of feature recognized in both of these two selected areas reaches the second preset ratio, we define these two selected areas as the combined selected area of this type of feature. That is to say, the combined selected area of this type of feature includes two areas. When actually performing recognition, it is necessary to recognize this type of feature in both positions of these two areas to possibly achieve recognition.

[0039] The advantage of this setting is that, on the one hand, it avoids the loss of the recognition function of this type of feature, makes up for the defect caused by the position setting of the camera 6, and thus can achieve the recognition of this type of feature to a certain extent. On the other hand, by using the selected areas of other types of features to define the combined selected area of this type of feature, it can effectively utilize the particularity of the selected areas of other types of features. Normally, the recognition degree of places that can be selected as selected areas is relatively high. By using this characteristic, it is convenient for the recognition of this type of feature and can effectively reduce the problems of increased computational complexity and low recognition accuracy caused by blind recognition in the entire image area.

[0040] Furthermore, the AI analysis box 5 is also used to re-select the combined selected area of this type of feature when the selection of the combined selected area of this type of feature fails.

[0041] Specifically, the second preset time period is set to 0.5 - 1 s, and the second preset ratio is set to 0.6.

[0042] Furthermore, the AI analysis box 5 is also used to receive the image information collected by the camera 6, perform feature recognition on the people in the combined selected area of this type of feature in the image. If this type of feature cannot be recognized in both of the two areas of the combined selected area of this type of feature, an alarm signal is sent to the terminal, that is, a normal risk recognition process is realized.

[0043] Refer to Figure 3 As shown, the square and hexagonal areas in the image are the selected areas of different types of features. After the above process, they are selected as the combined selected area of this type of feature. Then, when recognizing, it is required to achieve this type of feature in both positions of the two areas. If it is not recognized, it means there is a risk.

[0044] Regarding the issue of how to define the same recognition target in two regions, the target identity can be defined through face recognition or action recognition, so as to achieve the identity recognition of the target in the two regions.

[0045] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising..." or "including..." does not exclude the existence of additional elements in the process, method, article or terminal device comprising the said element. In addition, in this article, "greater than", "less than", "exceeding", etc. are understood not to include the present number; "above", "below", "within", etc. are understood to include the present number.

[0046] The above description of the embodiments is for the convenience of those of ordinary skill in the art to understand and use the present invention. Obviously, those who are familiar with the technology in the art can easily make various modifications to the embodiments and apply the general principles described herein to other embodiments without creative labor. Therefore, the present invention is not limited to the above embodiments, and the improvements and modifications made by those skilled in the art without departing from the scope of the present invention according to the disclosure of the present invention should be within the protection scope of the present invention.

Claims

1. An intelligent identification platform for highway construction safety risks based on artificial intelligence, characterized in that, Including: A terminal, which is used to present information after receiving risk identification information sent by a server; A server, which is used to receive risk identification information from an AI analysis box, and analyze, store, and transmit the risk identification information; An AI analysis box, which is used to receive image information collected by a camera, and analyze and process the image information. The process of analyzing and processing the image includes: receiving multiple images within a first preset time period, randomly selecting multiple regions in one of the images, defining them as candidate regions, and assigning the candidate regions to other images correspondingly. For each image, perform feature recognition of the same type of features within the candidate regions, define the candidate regions where the same type of features are recognized as preselected regions, count the number of preselected regions of multiple images according to the region positions, and define the region position with the largest number of preselected regions and exceeding the first preset ratio as the selected region of the same type of features. The first preset ratio refers to the ratio of the number of preselected regions in a certain region position preset in advance to the total number of images; A camera, which is used to be installed at a construction site to collect image information.

2. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, wherein The terminal includes a PC terminal, a mobile phone terminal, and a large screen terminal.

3. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, characterized in that, There are multiple AI analysis boxes and multiple cameras.

4. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, characterized in that, The first preset time period is set to 0.5 - 1 s, and the first preset ratio is set to 0.8 - 0.

9.

5. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, characterized in that, The same type of features are helmet wearing features, safety belt wearing features, or construction platform rest features.

6. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, characterized in that, The AI analysis box is also used to regularly clear the information of the selected region of the same type of features and then reselect.

7. An intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 1, characterized in that, The AI analysis box is also used to receive image information collected by the camera, and perform feature recognition on the people in the selected region of the same type of features in the image. If the same type of features cannot be recognized, an alarm signal is sent to the terminal.

8. An intelligent identification platform for highway construction safety risks based on artificial intelligence according to any one of claims 1 to 7, characterized in that, The AI analysis box is also used to reselect the selected region of the same type of features when the selection of the selected region of the same type of features fails.

9. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 8, characterized in that, The AI analysis box is also used to, when the selection of the jointly selected region of the same type of features fails, determine whether there are more than two selected regions of other feature types. If so, randomly select two selected regions of other feature types, and receive multiple images within a second preset time period collected by the camera, recognize the same type of features within the two selected regions of other feature types selected in each image, and count the number of images in which the same type of features are recognized within the two selected regions of other feature types selected. If, for each of the two selected regions of other feature types selected, the number of images in which the same type of features are recognized within each selected region of other feature types reaches the second preset ratio, then define these two selected regions of other feature types as the jointly selected region of the same type of features.

10. The intelligent identification platform for highway construction safety risks based on artificial intelligence according to claim 9, characterized in that, The AI analysis box is also used to reselect the jointly selected region of the same type of features when the selection of the jointly selected region of the same type of features fails.

Images