A method, apparatus, device and storage medium for automatically arranging a shower head

Abstract

The application discloses a method, device, equipment and storage medium for automatically arranging a spray head, and comprises the following steps: determining an arrangement area requiring the spray head from a target building model; performing three-dimensional grid sampling on the arrangement area to obtain a plurality of spatial feature points, and determining a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector; inputting the feature vector of each spatial feature point into a pre-trained prediction model, and obtaining a spray head arrangement probability of each spatial feature point output by the prediction model; and screening arrangement points from all the spatial feature points based on the spray head arrangement probability of each spatial feature point, and arranging the spray head at each arrangement point. The application can realize precise and efficient automatic arrangement of the spray head, and significantly reduce the dependence on manual design and model accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer-aided design, and in particular to a method, device and equipment for automatically arranging spray heads and a storage medium. BACKGROUND

[0002] In the field of building fire engineering design, the reasonable arrangement of spray heads is a key link to ensure fire safety. At present, the industry still relies on designers to manually position and arrange spray heads in building information models according to design specifications; this method is not only inefficient and tedious, but also highly dependent on the personal experience of designers, consumes a large amount of labor costs, and is difficult to ensure the consistency of design speed and quality. In order to improve efficiency, some software tools with automatic arrangement function have appeared on the market. However, these existing tools are mostly based on fixed geometric algorithms, and their applicability has significant limitations; they can usually only handle regular quadrilateral areas such as rectangles and trapezoids, and once they face complex polygons, arcs or irregular building spaces, their arrangement effect is often unsatisfactory, and they may even fail to complete the automatic arrangement. In addition, even if the arrangement is successful, the accuracy of the result is poor, and there are often problems such as unreasonable point positions and substandard spacing, which still require designers to make a lot of manual adjustments and corrections.

[0003] Therefore, how to overcome the deficiencies of the prior art and provide a spray head automatic arrangement scheme that can intelligently adapt to different spatial forms, has high arrangement accuracy and can significantly improve design efficiency has become a technical problem that technicians in the field need to solve urgently. SUMMARY

[0004] The purpose of the present application is to provide a method, device, equipment and storage medium for automatically arranging spray heads, which can realize accurate and efficient full-automatic arrangement of spray heads and significantly reduce the dependence on manual design and model accuracy.

[0005] According to one aspect of the present application, a method for automatically arranging spray heads is provided, the method comprising:

[0006] determining an arrangement area in which spray heads need to be arranged from a target building model;

[0007] performing three-dimensional grid sampling on the arrangement area to obtain a plurality of spatial feature points, and determining a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector;

[0008] inputting the feature vector of each spatial feature point into a pre-trained prediction model, and obtaining a spray head arrangement probability of each spatial feature point output by the prediction model;

[0009] screening, from all the spatial feature points, a layout point based on a layout probability of each spatial feature point, and arranging a sprinkler head at each layout point.

[0010] Optionally, the three-dimensional grid sampling of the layout area to obtain the spatial feature points comprises:

[0011] obtaining a space contour surrounded by wall elements and column elements in the layout area;

[0012] determining a three-dimensional bounding box of the space contour;

[0013] performing grid division along the X-axis, Y-axis and Z-axis directions at a preset sampling interval in the three-dimensional bounding box, and taking a grid intersection point as the spatial feature point.

[0014] Optionally, the determining of the feature value of each spatial feature point under a preset feature item to form a corresponding feature vector comprises:

[0015] determining a first feature value for representing whether the spatial feature point is in a sprinkler head prohibited layout area;

[0016] determining a second feature value for representing whether there is an obstacle around the spatial feature point that hinders installation and use of the sprinkler head;

[0017] determining a third feature value for representing a fire risk level of a region where the spatial feature point is located;

[0018] combining the first feature value, the second feature value and the third feature value into a feature vector of the spatial feature point.

[0019] Optionally, the screening, from all the spatial feature points, a layout point based on a layout probability of each spatial feature point, and arranging a sprinkler head at each layout point comprises:

[0020] screening, as a candidate point, a spatial feature point with a sprinkler head layout probability greater than a preset probability threshold;

[0021] filtering out, from all the candidate points, a point located outside an actual contour of the layout area to obtain a layout point.

[0022] Optionally, before the determining of the layout area in which the sprinkler head needs to be arranged from the target building model, the method further comprises:

[0023] obtaining a plurality of historical building models in which sprinkler heads have been correctly arranged;

[0024] extracting, from each historical building model, a plurality of sample space regions surrounded by wall elements and column elements;

[0025] respectively, and determine a feature vector of each sample feature point and a label for representing whether a sprinkler head is actually arranged at the sample feature point;

[0026] train a preset neural network model by taking the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model.

[0027] Optionally, the training of the preset neural network model by taking the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model, comprises:

[0028] In the process of training the neural network model, the model parameters are optimized by minimizing a composite loss function;

[0029] The composite loss function is obtained by weighted combination of the following three parts: a basic loss part for calculating the difference between the predicted sprinkler head arrangement probability of the model and the corresponding label; a distance loss part for calculating the spatial distance between the predicted sprinkler head position and the actual sprinkler head position; and a quantity loss part for calculating the absolute difference between the predicted number of sprinkler heads and the actual number of sprinkler heads.

[0030] To achieve the above-mentioned purpose, the application further provides an automatic sprinkler head arrangement device, which comprises:

[0031] A determination module is configured to determine an arrangement region requiring arrangement of a sprinkler head from a target building model;

[0032] A sampling module is configured to perform three-dimensional grid sampling on the arrangement region to obtain a plurality of spatial feature points, and determine a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector;

[0033] A prediction module is configured to input the feature vector of each spatial feature point into a pre-trained prediction model, and obtain a sprinkler head arrangement probability of each spatial feature point output by the prediction model;

[0034] An arrangement module is configured to screen arrangement points from all spatial feature points based on the sprinkler head arrangement probability of each spatial feature point, and arrange a sprinkler head at each arrangement point.

[0035] Optionally, the device further comprises a training module configured to:

[0036] obtain a plurality of historical building models in which sprinkler heads have been correctly arranged;

[0037] extract a plurality of sample space regions surrounded by wall elements and column elements from each historical building model, respectively;

[0038] perform three-dimensional grid sampling on each sample space region respectively to obtain a plurality of sample feature points, and determine a feature vector of each sample feature point and a label for representing whether a sprinkler head is actually arranged at the sample feature point;

[0039] training a preset neural network model by taking the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model.

[0040] To achieve the above object, the present application further provides a computer device, which specifically comprises a memory, a processor, and a computer program stored on the memory and capable of running on the processor, and the processor implements the steps of the above-mentioned method for automatically arranging sprinkler heads when executing the computer program.

[0041] To achieve the above object, the present application further provides a computer readable storage medium having a computer program stored thereon, and the computer program implements the steps of the above-mentioned method for automatically arranging sprinkler heads when executed by a processor.

[0042] The method, device, equipment and storage medium for automatically arranging sprinkler heads provided by the present application can accurately process various complex irregularly shaped regions by converting the arrangement problem into an intelligent decision based on three-dimensional grid points, greatly widening the application range of the automatic arrangement function. In addition, the model can be trained by using a deep learning algorithm, so that the model can deeply learn and internalize the fire arrangement specification, significantly improving the accuracy and rationality of the arrangement scheme, and the output result is more in line with the actual engineering requirements. The present application can realize accurate and efficient full-automatic arrangement of sprinkler heads, and significantly reduce the dependence on manual design and model accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0043] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments, and are not meant to limit the present application. Moreover, the same reference numerals are used throughout the various drawings to designate the same or similar parts. In the drawings:

[0044] Figure 1 An optional flowchart of the method for automatically arranging sprinkler heads provided for Example One;

[0045] Figure 2 A schematic diagram of a parameter setting interface provided for Example One;

[0046] Figure 3 A complete flowchart of the method for automatically arranging sprinkler heads based on deep learning provided for Example One;

[0047] Figure 4 The result schematic diagram of automatically arranging the spray head provided for the embodiment one is shown in the figure;

[0048] Figure 5 The flowchart of the preprocessing process of the historical sample data provided for the embodiment one is shown in the figure;

[0049] Figure 6 The visual display schematic diagram of the historical sample data provided for the embodiment one is shown in the figure;

[0050] Figure 7 The complete flowchart of the model training provided for the embodiment one is shown in the figure;

[0051] Figure 8 The optional component structure schematic diagram of the device for automatically arranging the spray head provided for the embodiment two is shown in the figure;

[0052] Figure 9 The optional hardware structure schematic diagram of the computer device provided for the embodiment three is shown in the figure. DETAILED DESCRIPTION

[0053] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] Embodiment one

[0055] The embodiment of the present application provides a method for automatically arranging a spray head, as shown in the figure, the method specifically comprises the following steps: Figure 1

[0056] Step S101: determining a layout area requiring a spray head from a target building model.

[0057] The target building model can be a BIM (Building Information Modeling, building information model); the layout area is a three-dimensional space, and the layout area can be a space area selected by a user from the target building model, or a closed area surrounded by wall elements and column elements automatically identified from the target building model.

[0058] Step S102: performing three-dimensional grid sampling on the layout area to obtain a plurality of space feature points, and determining a feature value of each space feature point under a preset feature item to form a corresponding feature vector.

[0059] ​In the embodiment, the arrangement region can be grid-sampled according to a preset sampling interval to obtain a plurality of spatial feature points; in addition, the feature vector of the spatial feature point can include at least one of the following: a prohibited arrangement feature for representing whether the spatial feature point is located in a prohibited arrangement region of the spray head, an obstacle feature for representing whether there is an obstacle around the spatial feature point which hinders the installation and use of the spray head, and a risk level feature for representing the fire risk level of the arrangement region; of course, in actual application, the spatial feature point can also include other dimensional features, such as a type feature for representing the type of the spray head to be arranged, a flow feature for representing the flow coefficient of the spray head, and the like, which are not limited herein.

[0060] Step S103: inputting the feature vector of each spatial feature point into the pre-trained prediction model, and obtaining the spray head arrangement probability of each spatial feature point output by the prediction model.

[0061] In the embodiment, the specified neural network model can be trained in advance by using historical sample data to obtain a prediction model; the input of the prediction model is the feature vector of each spatial feature point in the arrangement region, and the output of the prediction model is the spray head arrangement probability corresponding to each spatial feature point; preferably, the value of the spray head arrangement probability is between 0 and 1, the closer the spray head arrangement probability is to 0, the smaller the probability of arranging the spray head at the spatial feature point, and the closer the spray head arrangement probability is to 1, the greater the probability of arranging the spray head at the spatial feature point.

[0062] Step S104: based on the spray head arrangement probability of each spatial feature point, screening arrangement points from all spatial feature points, and arranging spray heads at each arrangement point.

[0063] Preferably, the spatial feature point with a spray head arrangement probability greater than a preset probability threshold can be set as an arrangement point.

[0064] In the embodiment, by converting the arrangement problem into an intelligent decision based on three-dimensional grid points, various complex irregularly shaped regions can be accurately processed, greatly widening the application range of the automatic arrangement function. In addition, by training the model by using a deep learning algorithm, the model can deeply learn and internalize the fire arrangement specification, significantly improving the accuracy and rationality of the arrangement scheme, and the output result is more in line with the actual engineering requirements. The embodiment can realize precise and efficient full-automatic arrangement of the spray head, and significantly reduce the dependence on manual design and model accuracy.

[0065] Specifically, the three-dimensional grid sampling of the arrangement region in step S102 to obtain a plurality of spatial feature points includes:

[0066] Step A1: obtaining a space contour surrounded by wall elements and column elements in the arrangement region;

[0067] Step A2: determining a three-dimensional bounding box of the space contour;

[0068] Step A3: dividing a grid in the three-dimensional bounding box along the X-axis, Y-axis and Z-axis directions at a preset sampling interval, and taking the grid intersection points as the space feature points.

[0069] Preferably, the preset sampling interval is 200 mm.

[0070] Further, the determination of the feature value of each space feature point under the preset feature item in the step S102 to form the corresponding feature vector specifically comprises:

[0071] Step A4: determining a first feature value for representing whether the space feature point is in a prohibited arrangement region of the spray head;

[0072] Wherein, the arrangement prohibited region is a region where the spray head should not be arranged according to the specification or experience; in the feature vector, a binary value is used to identify whether the space feature point is located in the prohibited region; preferably, it is judged whether the distance between the space feature point and any wall element is less than 100 mm, if yes, the first feature value is set to 1, and if not, the first feature value is set to 0.

[0073] Step A5: determining a second feature value for representing whether there is an obstacle around the space feature point which hinders the installation and use of the spray head;

[0074] Wherein, the obstacle refers to a building structure such as a wall, beam, column which will physically block the installation of the spray head or affect its spraying range in the arrangement region; in the feature vector, a binary value is used to represent; preferably, it is judged whether there is a wall element, beam element or column element at the space feature point, if yes, the second feature value is set to 1, and if not, the second feature value is set to 0.

[0075] Step A6: determining a third feature value for representing the fire risk level of the region where the space feature point is located;

[0076] Preferably, as shown in Figure 2 the parameter setting interface, the user can input the arrangement algorithm, arrangement region and arrangement parameters in the interface; wherein, the arrangement parameters include the danger level (i.e. the fire risk level) of the arrangement region; the fire risk level set by the user in advance for the arrangement region is obtained, and the third feature value is set to the fire risk level; wherein, the third feature value is represented by an integer type.

[0077] Step A7: Combine the first feature value, the second feature value, and the third feature value to form the feature vector of the spatial feature point.

[0078] In this embodiment, steps A1 to A7 above can discretize the complex building space into standard three-dimensional grid points, and assign a feature vector with a unified structure to each point, effectively solving the core technical problem that unstructured BIM model data cannot be directly processed by deep learning models. This preprocessing process digitizes information on components such as walls and columns, as well as design specifications (e.g., no-go zones within 100mm of a wall), into feature vectors containing obstacles, no-go zones, and hazard levels, providing standardized input for the neural network, enabling it to consider multi-dimensional spatial constraints simultaneously.

[0079] Furthermore, step S104, which involves selecting sprinkler head placement locations from all spatial feature points based on the sprinkler head placement probability for each spatial feature point, and placing sprinkler heads at each placement location, specifically includes:

[0080] Step B1: Select spatial feature points with a sprinkler head placement probability greater than a preset probability threshold as candidate locations;

[0081] Step B2: Filter out the points located outside the actual outline of the arrangement area from all candidate points to obtain the arrangement points.

[0082] In this embodiment, since the actual layout area may have various shapes, when a polygonal layout area is obtained, the bounding box of the polygonal area is first obtained, and then feature points are sampled on the bounding box at intervals of 200mm. However, this causes a problem: many points located outside the polygon are also collected. From a business logic perspective, these points should not be used to place sprinkler heads. Therefore, after obtaining the layout result of the model, these points should be filtered to improve the final display effect.

[0083] like Figure 3 The diagram shows the complete flowchart of an automated sprinkler head deployment system based on deep learning. The designer selects the deployment area and sets the deployment parameters. The system first performs grid sampling on the deployment area to obtain multiple feature points, and then inputs the feature vector of each feature point into a trained neural network prediction model to output the prediction result. The system deploys the sprinkler heads based on the prediction result. The designer can also fine-tune the automatically deployed sprinkler heads to meet business requirements. Figure 4 The image shown is a schematic diagram illustrating the result of automatically deploying sprinkler heads. Figure 4 The circles in the diagram represent the locations where the points are arranged.

[0084] Furthermore, prior to step S103, the method further includes:

[0085] Step C1: Obtain multiple historical building models with correctly arranged sprinkler heads.

[0086] Step C2: Extract multiple sample space regions enclosed by wall elements and column elements from each historical building model;

[0087] In this embodiment, it is necessary to collect a large number of historical projects in which the sprinkler heads have been arranged as required. A historical project may contain multiple spatial areas, and data will be collected on a spatial area basis, that is, a spatial area will be used as a sample.

[0088] Step C3: Perform three-dimensional mesh sampling on each sample space region to obtain multiple sample feature points, and determine the feature vector of each sample feature point and the label used to characterize whether a sprinkler head is actually placed at the sample feature point;

[0089] In this embodiment, a historical project is stored in the form of a BIM model. Such data cannot be directly learned by a neural network, so a preprocessing algorithm is needed to extract useful information from the BIM model. The processing logic of the preprocessing algorithm is as follows: Figure 5 As shown, firstly, the polygonal outline enclosed by walls and columns in the space is obtained on a spatial basis, and the bounding box of the polygonal outline is generated; then, sample feature points are collected at 200mm intervals in the X, Y, and Z axes; then, for each sample feature point, its feature vector is calculated, which contains the following three dimensions: 1) Restricted area layout feature: Determine whether the sample feature point is less than 100mm away from any wall, if so, it is 1, otherwise it is 0; 2) Obstacle feature: Determine whether the coordinates of the sample feature point contain walls, beams, or columns, if so, it is 1, otherwise it is 0; 3) Risk level feature: Determined according to the fire risk level of the area where the sample feature point is located, and represented as an integer; finally, a corresponding label needs to be generated for each sample feature point. When the coordinates of the sample feature point contain a sprinkler head, the label is 1, otherwise it is 0. It should also be noted that the "restricted area layout feature" mentioned above is equivalent to the "first feature value", the "obstacle feature" is equivalent to the "second feature value", and the "risk level feature" is equivalent to the "third feature value". It should be noted that the higher the fire risk level, the denser the sprinkler head arrangement should be. In order for the model to adapt to the arrangement requirements of different fire risk levels, this information needs to be input into the model so that the model can learn the relationship between sprinkler head spacing and fire risk level.

[0090] Additionally, the feature extraction results of all sample feature points are serialized into a JSON file for storage, resulting in a dataset; the visualization of the JSON file content is shown below. Figure 6 As shown, in Figure 6In the figure, the red lines represent the area where the obstacle exists, the blue lines represent the area where the sprinkler is prohibited to be arranged, the yellow points represent the actual arrangement points of the sprinkler, and the gray represents the fire risk level (the higher the risk level, the darker the gray color).

[0091] Step C4: training a preset neural network model by taking the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model.

[0092] Preferably, the preset neural network model is a CNN (Convolutional Neural Network) model.

[0093] In the embodiment, the model training process represented by the above steps C1 to C4 can convert a large number of historical BIM models with correctly arranged sprinklers into a standardized grid feature point dataset, solving the technical problems of lack of high-quality training data for deep learning models and inability to directly learn complex arrangement rules from unstructured BIM models. By accurately generating a feature vector containing arrangement exclusion zones, obstacles, and risk levels and corresponding labels for each sample feature point, the designer's experience and fire safety specifications are digitized into model-learnable objectives. This process enables the CNN model to automatically induce the internal rules of sprinkler arrangement from the data, rather than relying on fixed rules written by humans. The finally trained model can intelligently adapt to different spatial forms, achieve precise and standardized automatic arrangement, significantly reduce the dependence of the algorithm on specific spatial shapes, and reduce the workload of manual adjustment.

[0094] Further, in the step C4, it further specifically includes:

[0095] In the process of training the neural network model, the model parameters are optimized by minimizing the composite loss function.

[0096] The composite loss function is obtained by weighted combination of the following three parts: a basic loss part for calculating the difference between the predicted sprinkler arrangement probability of the model and the corresponding label; a distance loss part for calculating the spatial distance between the predicted sprinkler position and the actual sprinkler position; and a quantity loss part for calculating the absolute difference between the predicted number of sprinklers and the actual number of sprinklers.

[0097] In this embodiment, the performance of the current model can be evaluated using a loss function during training. A smaller loss indicates a better model performance, and vice versa. Simultaneously, the model is also optimized during training based on the calculation results of the loss function. Therefore, the design of the loss function is crucial to the final result. Preferably, the loss function used in this embodiment includes the following three parts: 1) Base loss: calculating the difference between the model's predicted result and the actual result using the cross-entropy function; 2) Distance loss: calculating the distance between the predicted sprinkler heads and the actual sprinkler heads, using the distance as the loss; the larger the distance, the larger the loss; as the distance between the predicted and actual sprinkler heads gradually decreases, the loss gradually decreases; 3) Quantity loss: calculating the absolute value of the predicted and actual sprinkler head quantities; the larger the absolute value, the larger the loss; otherwise, the loss gradually decreases.

[0098] In this embodiment, after designing the neural network model and loss function, multiple rounds of training and validation are performed using the sample dataset to obtain a predictive model that has learned the layout specifications.

[0099] In addition, the diversity of sample data is an important factor in ensuring the training effect. Therefore, the acquisition of multiple historical building models with correctly arranged sprinkler heads in step C1 specifically includes: rotating the acquired historical model along the Z-axis at multiple angles to obtain multiple sample building models, and then performing the above steps C2 to C4 based on all the obtained sample building models.

[0100] like Figure 7 The diagram shows the complete process of training a deep learning-based model. In the model training process, historical items are first collected, and then features are extracted from the historical items in spatial units to generate a training sample dataset. The preset model is then trained using a preset model and a constructed composite loss function to obtain the final prediction model.

[0101] In this embodiment, the building space is discretized into standard three-dimensional grid points, and each point is assigned a unified structured feature vector to achieve a digital representation of the layout environment. During model training, a composite loss function is used to guide the neural network model to optimize the accuracy of the sprinkler head positions and the rationality of their number. To obtain better training results, the diversity of training data is a crucial factor affecting the training effect. The extracted training model is randomly rotated, and the rotated model is used as new training input to enrich the sample data.

[0102] Example 2

[0103] This invention provides an automatic sprinkler head arrangement device, such as... Figure 8 As shown, the device specifically includes the following components:

[0104] The determination module 801 is configured to determine a layout area requiring arrangement of a sprinkler head from a target building model;

[0105] The sampling module 802 is configured to perform three-dimensional grid sampling on the layout area to obtain a plurality of spatial feature points, and determine a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector;

[0106] The prediction module 803 is configured to input the feature vector of each spatial feature point into a pre-trained prediction model, and obtain a sprinkler head arrangement probability of each spatial feature point output by the prediction model;

[0107] The arrangement module 804 is configured to screen arrangement points from all spatial feature points based on the sprinkler head arrangement probability of each spatial feature point, and arrange a sprinkler head at each arrangement point.

[0108] Specifically, the sampling module 802 is configured to:

[0109] Obtain a space contour surrounded by wall elements and column elements in the layout area;

[0110] Determine a three-dimensional bounding box of the space contour;

[0111] Divide the three-dimensional bounding box into grids along the X-axis, Y-axis and Z-axis directions at a preset sampling interval, and take the grid intersection points as the spatial feature points.

[0112] Further, the sampling module 802 is further configured to:

[0113] Determine a first feature value for representing whether the spatial feature point is in a sprinkler head prohibited arrangement area;

[0114] Determine a second feature value for representing whether there is an obstacle around the spatial feature point that hinders installation and use of the sprinkler head;

[0115] Determine a third feature value for representing a fire risk level of the area where the spatial feature point is located;

[0116] Combine the first feature value, the second feature value and the third feature value into the feature vector of the spatial feature point.

[0117] Further, the arrangement module 804 is specifically configured to:

[0118] Screen spatial feature points with a sprinkler head arrangement probability greater than a preset probability threshold as candidate points;

[0119] Filter out the point positions located outside the actual contour of the arrangement region from all candidate point positions to obtain arrangement point positions.

[0120] In addition, the device further comprises a training module configured to:

[0121] obtain a plurality of historical building models in which the spray heads have been correctly arranged;

[0122] extract a plurality of sample space regions surrounded by wall elements and column elements from each historical building model, respectively;

[0123] perform three-dimensional grid sampling on each sample space region to obtain a plurality of sample feature points, and determine a feature vector of each sample feature point and a label for representing whether a spray head is actually arranged at the sample feature point;

[0124] train a preset neural network model by taking the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model.

[0125] Further, the training module is specifically configured to:

[0126] optimize the model parameters by minimizing a composite loss function during the training of the neural network model;

[0127] The composite loss function is obtained by weighted combination of the following three parts: a basic loss part for calculating the difference between the predicted spray head arrangement probability of the model and the corresponding label; a distance loss part for calculating the spatial distance between the predicted spray head position and the actual spray head position; and a quantity loss part for calculating the absolute difference between the predicted number of spray heads and the actual number of spray heads.

[0128] Embodiment Three

[0129] The embodiment also provides a computer device, such as a smartphone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server, or a cabinet server (including a standalone server or a server cluster composed of multiple servers), which can execute programs. As shown in the figure, the computer device 90 of the embodiment at least includes but is not limited to a memory 901 and a processor 902 which can be connected to each other through a system bus. It should be noted that, Figure 9 Only the computer device 90 with components 901-902 is shown, but it should be understood that all the shown components are not required to be implemented, and more or fewer components can be alternatively implemented. Figure 9

[0130] ​In this embodiment, the memory 901 (i.e., a readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 901 can be an internal storage unit of the computer device 90, such as a hard disk or a memory of the computer device 90. In other embodiments, the memory 901 can also be an external storage device of the computer device 90, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 90. Of course, the memory 901 can also include both the internal storage unit and the external storage device of the computer device 90. In this embodiment, the memory 901 is generally used to store an operating system and various application software installed on the computer device 90. In addition, the memory 901 can also be used to temporarily store various data that have been output or will be output.

[0131] The processor 902 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 902 is generally used to control the overall operation of the computer device 90.

[0132] Specifically, in this embodiment, the processor 902 is used to execute a program of the method for automatically arranging a sprinkler head stored in the memory 901, and the program of the method for automatically arranging a sprinkler head is implemented when executed as follows:

[0133] determining an arrangement area requiring arrangement of a sprinkler head from a target building model;

[0134] performing three-dimensional grid sampling on the arrangement area to obtain a plurality of spatial feature points, and determining a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector;

[0135] inputting the feature vector of each spatial feature point into a pre-trained prediction model, and obtaining a sprinkler head arrangement probability of each spatial feature point output by the prediction model;

[0136] based on the sprinkler head arrangement probability of each spatial feature point, screening arrangement point positions from all spatial feature points, and arranging a sprinkler head at each arrangement point position.

[0137] The specific embodiment process of the above method steps can be seen from the first embodiment, which will not be repeated here.

[0138] Embodiment four

[0139] The embodiment also provides a computer readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card memory (for example, an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application market, etc., which stores a computer program, and the computer program is executed by a processor to implement the following method steps:

[0140] Determining a layout area requiring to arrange a sprinkler head from the target building model;

[0141] Carrying out three-dimensional grid sampling on the layout area to obtain a plurality of spatial feature points, and determining a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector;

[0142] Inputting the feature vector of each spatial feature point into a pre-trained prediction model, and obtaining a sprinkler head arrangement probability of each spatial feature point output by the prediction model;

[0143] Based on the sprinkler head arrangement probability of each spatial feature point, screening out arrangement points from all spatial feature points, and arranging a sprinkler head at each arrangement point.

[0144] The specific embodiment process of the above method steps can be seen from the first embodiment, which will not be repeated here.

[0145] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the sentence "includes a…" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0146] The above embodiment numbers of the present application are only for description, not representing the advantages and disadvantages of the embodiments.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and the necessary general hardware platform, of course, can also be realized by hardware, but in many cases, the former is a better embodiment.

[0148] The above are only preferred embodiments of the present application, and do not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method of automatically arranging a showerhead, characterized by, The method comprises: determining a layout area requiring a sprinkler head from a target building model; performing three-dimensional grid sampling on the layout area to obtain a plurality of spatial feature points, and determining a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector; inputting the feature vector of each spatial feature point into a pre-trained prediction model, and obtaining a sprinkler head layout probability of each spatial feature point output by the prediction model; screening layout points from all spatial feature points based on the sprinkler head layout probability of each spatial feature point, and laying a sprinkler head at each layout point; wherein the determination of the feature value of each spatial feature point under the preset feature item to form the corresponding feature vector comprises: determining a first feature value for representing whether the spatial feature point is in a sprinkler head prohibited layout area; determining a second feature value for representing whether there is an obstacle around the spatial feature point that hinders the installation and use of the sprinkler head; determining a third feature value for representing a fire risk level of the area where the spatial feature point is located; combining the first feature value, the second feature value and the third feature value into the feature vector of the spatial feature point.

2. The method of automatically arranging a spray head of claim 1, wherein, The three-dimensional grid sampling on the layout area to obtain a plurality of spatial feature points comprises: obtaining a space contour surrounded by wall elements and column elements in the layout area; determining a three-dimensional bounding box of the space contour; performing grid division in the three-dimensional bounding box along the X-axis, Y-axis and Z-axis directions at a preset sampling interval, and taking the grid intersection points as the spatial feature points.

3. The method of automatically arranging a spray head of claim 1, wherein, The screening of layout points from all spatial feature points based on the sprinkler head layout probability of each spatial feature point, and the laying of a sprinkler head at each layout point comprise: screening spatial feature points with a sprinkler head layout probability greater than a preset probability threshold as candidate points; filtering out points outside the actual contour of the layout area from all candidate points to obtain layout points.

4. The method of automatically arranging a showerhead according to any one of claims 1 to 3, wherein, Before the determination of the layout area requiring a sprinkler head from the target building model, the method further comprises: obtaining a plurality of historical building models with correctly laid sprinkler heads; extracting a plurality of sample space areas surrounded by wall elements and column elements from each historical building model respectively; performing three-dimensional grid sampling on each sample space area to obtain a plurality of sample feature points, and determining a feature vector of each sample feature point and a label for representing whether a sprinkler head is actually laid at the sample feature point; training a preset neural network model with the feature vector of each sample feature point as input and the label as expected output to obtain the prediction model.

5. The method of automatically arranging a showerhead of claim 4, wherein, The training of the preset neural network model with the feature vector of each sample feature point as input and the label as expected output to obtain the prediction model comprises: optimizing model parameters by minimizing a composite loss function in the process of training the neural network model; The composite loss function is obtained by weighted combination of three parts: a basic loss part for calculating the difference between the predicted spray head arrangement probability and the corresponding label; a distance loss part for calculating the spatial distance between the predicted spray head position and the actual spray head position; and a quantity loss part for calculating the absolute difference between the predicted spray head quantity and the actual spray head quantity.

6. An apparatus for automatically arranging a showerhead, characterized by, The device comprises: A determination module configured to determine an arrangement area requiring spray head arrangement from a target building model; A sampling module configured to perform three-dimensional grid sampling on the arrangement area to obtain a plurality of spatial feature points, and determine a feature value of each spatial feature point under a preset feature item to form a corresponding feature vector; A prediction module configured to input the feature vector of each spatial feature point into a pre-trained prediction model, and obtain a spray head arrangement probability of each spatial feature point output by the prediction model; An arrangement module configured to screen arrangement points from all spatial feature points based on the spray head arrangement probability of each spatial feature point, and arrange spray heads at each arrangement point; The sampling module is configured to: Determine a first feature value for representing whether the spatial feature point is in a spray head prohibited arrangement area; Determine a second feature value for representing whether there is an obstacle around the spatial feature point that hinders installation and use of the spray head; Determine a third feature value for representing a fire risk level of the area where the spatial feature point is located; Combine the first feature value, the second feature value and the third feature value into the feature vector of the spatial feature point.

7. The apparatus of claim 6, wherein, The device further comprises a training module configured to: Obtain a plurality of historical building models in which spray heads have been correctly arranged; Extract a plurality of sample space areas surrounded by wall elements and column elements from each historical building model, respectively; Perform three-dimensional grid sampling on each sample space area to obtain a plurality of sample feature points, and determine a feature vector of each sample feature point and a label for representing whether a spray head is actually arranged at the sample feature point; Train a preset neural network model with the feature vector of each sample feature point as input and the label as expected output, to obtain the prediction model.

8. A computer device comprising: A memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of any one of claims 1 to 5 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 5.

Images