Method and device for determining a placement position of an automated teller machine, and storage medium

Abstract

The application discloses a kind of automatic teller machine placement position determination method, device and storage medium.It relates to artificial intelligence technical field, wherein, the method includes: target image is input into target prediction model, and initial prediction map is output, wherein, target image is the image of the environment of automatic teller machine to be placed, and initial prediction map at least includes: the gaze point area information of the first class object predicted by target prediction model;The target image is calculated by vanishing point detection algorithm, and the priori graph of target image is determined, wherein, at least including: the position of vanishing point;Initial prediction map and priori graph are weighted and fused, and target prediction map is obtained, wherein, target prediction map is marked with the to-be-placed position of automatic teller machine in environment.The application solves the technical problem that manual selection mode determines the placement position of automatic teller machine (ATM), and users are not easy to find the position of ATM, leading to low utilization rate of ATM.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, and storage medium for determining the placement location of an ATM. Background Technology

[0002] Currently, in order to meet people's cash consumption needs, reflect the high-quality services of financial institutions, and promote the flow of funds, ATMs are usually placed in large shopping malls. Their placement is mainly determined by staff based on experience or the specific requirements of the mall. It is impossible to determine whether their location conforms to the behavioral habits of most people, and whether most people can quickly find them when they have cash consumption needs in the mall. There is a lack of scientific theoretical basis.

[0003] Therefore, the placement of ATMs in large shopping malls is mainly determined by staff based on experience or the mall's specific requirements, which leads to low efficiency in selecting ATM locations and low ATM utilization.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method, apparatus, and storage medium for determining the placement location of an ATM, thereby at least solving the technical problem of low ATM utilization caused by the difficulty for users to find the ATM location due to manual selection.

[0006] According to one aspect of the present invention, a method for determining the placement location of an ATM is provided, comprising: inputting a target image into a target prediction model and outputting an initial prediction map, wherein the target image is an image of the environment in which the ATM is to be placed, and the initial prediction map includes at least: gaze point region information of a first type of object predicted by the target prediction model; calculating a prior map of the target image using a vanishing point detection algorithm, wherein the prior map includes at least: the location of a vanishing point, the vanishing point being the vanishing point of a target sign in the target image; and performing a weighted fusion of the initial prediction map and the prior map to obtain a target prediction map, wherein the target prediction map marks the placement location of the ATM in the environment.

[0007] Further, the prior image of the target image is determined by calculating the vanishing point detection algorithm, including: extracting texture features from the target image using a target filter to obtain a first image; calculating the position of the vanishing point in the first image using a local voting algorithm; and performing Gaussian blur processing on the first image based on the position of the vanishing point and a preset Gaussian weight to obtain the prior image.

[0008] Furthermore, before extracting texture features from the target image using the target filter, the method includes: preprocessing the target image, wherein the preprocessing includes at least one of the following: grayscale conversion and noise reduction.

[0009] Further, the target prediction model is obtained by: acquiring an eye-tracking dataset, wherein the eye-tracking dataset includes at least: a target gaze map of the second type of object in the environment, the target gaze map including at least: gaze data of the second type of object searching for the ATM in the target image; and training the initial prediction model based on the eye-tracking dataset to obtain the target prediction model.

[0010] Furthermore, acquiring the eye-tracking dataset also includes: collecting eye-tracking data of the second type of object searching for the ATM in the target image; generating the target gaze map based on the eye-tracking data; and determining the eye-tracking dataset based on the target gaze map.

[0011] Further, generating the target gaze map based on the eye-tracking data includes: generating an initial gaze map of the target image based on the eye-tracking data; and performing a convolution operation on the initial gaze map using a target Gaussian function to obtain the target gaze map.

[0012] Further, the initial prediction map and the prior map are weighted and fused to obtain the target prediction map, including: inputting the target image into the target neural network model and outputting prior weights; and fusing the initial prediction map and the prior map using the prior weights to obtain the target prediction map.

[0013] Furthermore, the target neural network model is obtained by: acquiring a training set, wherein the training set includes at least: images of the target signboard that exist and images of the target signboard that do not exist; training an initial neural network model using the training set to obtain the target neural network model, wherein the type of the target neural network model includes at least: a binary classification neural network model.

[0014] According to one aspect of the present invention, an apparatus for determining the placement location of an ATM is provided, comprising: a processing unit, configured to input a target image into a target prediction model and output an initial prediction map, wherein the target image is an image of the environment in which the ATM is to be placed, and the initial prediction map includes at least: gaze point region information of a first type of object predicted by the target prediction model; a determining unit, configured to calculate a prior map of the target image using a vanishing point detection algorithm, wherein the prior map includes at least: the location of a vanishing point, the vanishing point being the vanishing point of a target sign in the target image; and a fusion unit, configured to perform weighted fusion of the initial prediction map and the prior map to obtain a target prediction map, wherein the target prediction map marks the placement location of the ATM in the environment.

[0015] Further, the determining unit includes: an extraction subunit, used to extract texture features from the target image using a target filter to obtain a first image; a calculation subunit, used to calculate the position of the vanishing point in the first image using a local voting algorithm; and a processing subunit, used to perform Gaussian blur processing on the first image based on the position of the vanishing point and a preset Gaussian weight to obtain the prior image.

[0016] Furthermore, the determining unit further includes: preprocessing the target image before extracting texture features from the target image using a target filter, wherein the preprocessing includes at least one of the following: grayscale conversion and noise reduction.

[0017] Further, the target prediction model is obtained through the following means: an acquisition unit, used to acquire an eye-tracking dataset, wherein the eye-tracking dataset includes at least: a target gaze map of the second type of object in the environment, the target gaze map including at least: gaze data of the second type of object searching for the ATM in the target image; and a training unit, used to train the initial prediction model based on the eye-tracking dataset to obtain the target prediction model.

[0018] Furthermore, the acquisition unit includes: an acquisition subunit for acquiring eye movement data of the second type of object searching for the ATM in the target image; a generation subunit for generating the target gaze map based on the eye movement data; and a determination subunit for determining the eye movement dataset based on the target gaze map.

[0019] Furthermore, the generation subunit includes: a generation module, used to generate an initial gaze point map of the target image based on the eye-tracking data; and a processing module, used to perform a convolution operation on the initial gaze point map using a target Gaussian function to obtain the target gaze point map.

[0020] Furthermore, the fusion unit includes: an input-output subunit, used to input the target image into the target neural network model and output prior weights; and a fusion processing subunit, used to perform fusion processing on the initial prediction map and the prior map through the prior weights to obtain the target prediction map.

[0021] Further, the target neural network model is obtained through the following methods: a training set acquisition subunit, used to acquire a training set, wherein the training set includes at least: images of the target signboard and images of the target signboard not existing; and a training subunit, used to train an initial neural network model using the training set to obtain the target neural network model, wherein the type of the target neural network model includes at least: a binary classification neural network model.

[0022] According to another aspect of the present invention, an electronic device is also provided, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method for determining the placement position of an ATM as described above by executing the executable instructions.

[0023] According to another aspect of the present invention, a computer-readable storage medium is also provided, which stores a computer program, wherein the computer program controls the device where the computer-readable storage medium is located to execute the method for determining the placement position of an ATM as described above when the computer program is running.

[0024] In this invention, a target image is input into a target prediction model, which outputs an initial prediction map. The target image is an image of the environment in which the ATM is to be placed. The initial prediction map includes at least the gaze point region information of the first type of object predicted by the target prediction model. A vanishing point detection algorithm is used to calculate a prior map of the target image, which includes at least the location of the vanishing point, which is the vanishing point of the target sign in the target image. The initial prediction map and the prior map are weighted and fused to obtain a target prediction map, which marks the location of the ATM in the environment. This solves the technical problem of low ATM utilization caused by the manual selection method for determining the ATM location, making it difficult for users to find the ATM. In this invention, an initial prediction map is output through an initial prediction model, and a prior map is determined through a vanishing point detection algorithm. The initial prediction map and the prior map are then weighted and fused to obtain a target prediction map. The placement location of the ATM is selected based on the target prediction map. This avoids the situation in related technologies where the placement location of the ATM is determined by manual selection, which has poor results. Thus, the technical effects of improving the selection efficiency of ATM placement location and improving ATM utilization are achieved. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 This is a flowchart of an optional method for determining the placement location of an ATM according to an embodiment of the present invention;

[0027] Figure 2 This is a flowchart of an optional method for determining a priori images of a target image according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of an optional device for determining the placement position of an ATM according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with relevant users or organizations. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent information from the aforementioned user or organization.

[0033] Example 1

[0034] According to an embodiment of the present invention, a method embodiment for determining the placement location of an optional ATM is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0035] Figure 1 This is a flowchart of an optional method for determining the placement location of an ATM according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0036] Step S101: Input the target image into the target prediction model and output the initial prediction map. The target image is an image of the environment in which the ATM is to be placed. The initial prediction map includes at least the gaze point region information of the first type of object predicted by the target prediction model.

[0037] The environment in which the ATM is to be placed can include, but is not limited to, indoor places such as large shopping malls. The first type of object can be a user who wants to use the ATM. The target prediction model can be a trained neural network model, such as SAM-ResNet (SAM: Self-Attention Enhancement Mechanism, ResNet Convolutional Neural Network Model). The initial prediction map can include the region information of the gaze point of the first type of object in the target image.

[0038] Step S102: Calculate the target image using a vanishing point detection algorithm to determine the prior image of the target image. The prior image includes at least the location of the vanishing point, which is the vanishing point of the target sign in the target image.

[0039] In this embodiment, the target image can be calculated using a vanishing point detection algorithm to obtain a priori map corresponding to the target image. The priori map may include the vanishing point of the target sign in the target image.

[0040] The aforementioned target signs may include, but are not limited to, road signs in the target image, such as signs for toilets, entrances, and exits. By determining the prior map of the target image, the visual attention region of the first type of object in the target image can be further predicted. The position and direction of the vanishing point obtained by detecting the sign are used as prior information about the location of the ATM in the environment where the ATM is to be placed. Based on this prior information, the prior map of the target image is determined.

[0041] Step S103: The initial prediction map and the prior map are weighted and fused to obtain the target prediction map, wherein the target prediction map marks the location where the ATM will be placed in the environment.

[0042] The target prediction map above can mark the location information of the ATM (Automated Teller Machine) to be placed. The location information may include, but is not limited to: the area information of the ATM placement location, the location coordinate information, and the reference objects of the ATM at the placement location. Based on the ATM's placement location in the environment marked in the target prediction map above, the ATM can be placed at the corresponding placement location. For example, the ATM can be placed at the placement location based on the ATM's location information, such as the floor, specific reference objects, area markers, room markers, etc. in the indoor environment.

[0043] To further improve the rationality of ATM placement in an environment where ATMs are to be placed, and to improve ATM utilization while facilitating the search for ATMs by the first type of target, a weighted fusion of the initial prediction map and the prior map can be performed to obtain the target prediction map. Based on the target prediction map, the placement location of the ATM can be selected.

[0044] By establishing an eye-tracking dataset based on the task of finding ATMs in a shopping mall, prior information is extracted from it and then combined with a gaze prediction model to construct a gaze prediction model driven by the task of finding ATMs in a shopping mall, so as to predict the reasonable placement of ATMs in the mall.

[0045] For an input image, an initial gaze prediction map (corresponding to the initial prediction map) can first be extracted using the SAM model (corresponding to the target prediction model mentioned above); at the same time, a prior map with prior information is obtained through the vanishing point detection algorithm, and then prior weights are obtained through a binary classification network. The initial prediction map and the prior map are then weighted and fused to obtain the final gaze prediction map (corresponding to the target prediction map mentioned above).

[0046] Through the above steps, in this embodiment, an initial prediction map is output through an initial prediction model, a prior map is determined through a vanishing point detection algorithm, and a weighted fusion of the initial prediction map and the prior map is performed to obtain a target prediction map. The placement location of the ATM is selected based on the target prediction map. This avoids the situation in related technologies where the placement of ATMs is determined manually, resulting in poor performance. Therefore, it achieves the technical effect of improving the efficiency of ATM placement selection and increasing ATM utilization. Furthermore, it solves the technical problem of low ATM utilization caused by the manual selection method, which makes it difficult for users to find the ATM location.

[0047] Figure 2 This is a flowchart of an optional method for determining a priori map of a target image according to an embodiment of the present invention. The method involves calculating the priori map of the target image using a vanishing point detection algorithm, including:

[0048] Step S201: Extract texture features from the target image using a target filter to obtain the first image;

[0049] Step S202: Calculate the position of the vanishing point in the first image using a local voting algorithm;

[0050] Step S203: Based on the location of the vanishing point and the preset Gaussian weights, perform Gaussian blur processing on the first image to obtain the prior image.

[0051] The target filters mentioned above may include, but are not limited to: Gabor filters (filters used to extract texture information in different directions).

[0052] To improve the accuracy of extracting prior information from the target image, a texture-based vanishing point detection method can be used. The vanishing point direction obtained by detecting the sign can be used as prior information about the ATM's location in the environment where it will be placed. The image is analyzed using a target filter to extract texture features (such as texture direction), resulting in a first image. After obtaining the texture direction of each pixel in the image, a voting algorithm is used to estimate the vanishing point's location. Then, a Gaussian blur is applied to the first image using preset Gaussian weights to generate a prior image.

[0053] Optionally, before extracting texture features from the target image using the target filter, the method includes: preprocessing the target image, wherein the preprocessing includes at least one of the following: grayscale conversion and denoising.

[0054] In this embodiment, the image of the environment in which the ATM is to be placed (corresponding to the target image mentioned above) can be preprocessed. The target image can be preprocessed through grayscale conversion, noise reduction, etc. The environment in which the ATM is to be placed can include, but is not limited to, shopping malls, thereby achieving the technical effect of improving image quality.

[0055] Optionally, the target prediction model is obtained by: acquiring an eye-tracking dataset, wherein the eye-tracking dataset includes at least: a target gaze map of the second type of object in the environment, and the target gaze map includes at least: gaze data of the second type of object searching for the ATM in the target image; and training the initial prediction model based on the eye-tracking dataset to obtain the target prediction model.

[0056] In this embodiment, the target prediction model can be obtained by training the SAM-ResNet, the best-performing fixation prediction model on standard datasets, using the aforementioned eye-tracking dataset. Alternatively, the target prediction model can be obtained by directly training the neural network model using the aforementioned eye-tracking dataset. This achieves the technical effect of improving the accuracy of the prediction results output by the target prediction model.

[0057] Optionally, acquiring the eye-tracking dataset may further include: collecting eye-tracking data of the second type of object searching for the ATM in the target image; generating a target gaze map based on the eye-tracking data; and determining the eye-tracking dataset based on the target gaze map.

[0058] In this embodiment, obtaining the eye-tracking dataset can begin by acquiring images of any scene within the environment of the ATM to be deployed. Then, an eye-tracking experiment is designed to find the ATM. Valid eye-tracking data is collected using a target device to generate a gaze point map. The target device may include an eye tracker. The generated gaze point map is processed using Gaussian weights to obtain a target gaze point map. The obtained target gaze point map constitutes the eye-tracking dataset for the ATM to be deployed environment, which is then used for training and testing of the gaze point prediction model (corresponding to the target prediction model mentioned above).

[0059] Specifically, 1,000 images collected in any environment where an ATM is to be deployed can be used as experimental materials. For example, if the environment where the ATM is to be deployed is a shopping mall, then 1,000 images collected in any shopping mall can be used as experimental materials.

[0060] Participants (corresponding to the second type of subjects mentioned above) were invited to view images of these experimental materials while the target device recorded their eye fixation data. Before each recording session, participants were asked to "find the ATM in this shopping mall," viewing the images with the task of "finding the ATM in this environment," thus simulating top-down selective attention. The eye-tracking records (corresponding to the eye-tracking data mentioned above) were formatted to obtain information such as fixation duration, X and Y coordinates of the fixation point, etc. These were then convolved using a two-dimensional Gaussian function to obtain the target fixation point map, achieving the technical effect of improving the data accuracy of the eye-tracking dataset.

[0061] Optionally, a target gaze map is generated based on eye-tracking data, including: generating an initial gaze map of the target image based on eye-tracking data; and performing a convolution operation on the initial gaze map using a target Gaussian function to obtain the target gaze map.

[0062] In this implementation, eye-tracking records (corresponding to the eye-tracking data mentioned above) can be organized according to a format to obtain information such as fixation duration, X and Y coordinates of the fixation point, and generate an initial fixation map. In the initial fixation map, the fixation point can be a very small area, similar to a point. In order to appropriately expand the area of ​​the fixation point to facilitate the prediction of the ATM machine's placement, a two-dimensional Gaussian function can be used for convolution to finally obtain the target fixation map, thereby achieving the technical effect of improving the data accuracy of the eye-tracking dataset.

[0063] Optionally, a weighted fusion of the initial prediction map and the prior map is performed to obtain the target prediction map, including: inputting the target image into the target neural network model and outputting prior weights; and fusing the initial prediction map and the prior map using the prior weights to obtain the target prediction map.

[0064] By inputting the target image into the aforementioned target neural network model, the output value is a measure of the prior tendency of the binary classification neural network model for the input target image. The larger the measure value, the greater the role of prior information. The output measure value is used as the prior weight to fuse the initial prediction map and the prior map to obtain the target prediction map. This achieves the technical effect of improving the rationality of the predicted ATM placement and improving the utilization rate of ATMs.

[0065] Optionally, the target neural network model is obtained by: acquiring a training set, wherein the training set includes at least: images of target signs and images of target signs that do not exist; training the initial neural network model using the training set to obtain the target neural network model, wherein the type of the target neural network model includes at least: a binary classification neural network model.

[0066] To measure the prior tendency of the vanishing point of the sign in the input image, a VGG-16 binary classification neural network model (corresponding to the target neural network model mentioned above) can be trained in this embodiment.

[0067] In this embodiment, the images in the training set can be divided into two categories: images with meaningful prior information (images containing target signs) and images without meaningful prior information (images without target signs). During model training, the Sigmoid activation function can be used in the neural network model, with cross-entropy as the loss function. The convergence of the initial neural network model is determined by calculating cross-entropy. After convergence, the target neural network model is obtained. By inputting the target image into the target neural network model, the output value is a measure of the prior tendency of the binary classification neural network model towards the input target image. A larger value indicates a greater influence of prior information, and this output measure is used as the prior weight.

[0068] It should be noted that this embodiment can be applied to the scenario of finding an ATM in a shopping mall. The neural network model combines the top-down attention mechanism and the bottom-up attention mechanism in the task of finding an ATM, and the detection results are more consistent with the visual attention mechanism and most people's perception of the task of finding an ATM.

[0069] In this embodiment, the gaze prediction model for the task of finding ATMs in a shopping mall consists of three parts: a basic gaze prediction model (corresponding to the initial prediction model mentioned above), prior information, and prior fusion. The basic gaze prediction module is based on SAM-ResNet, the current state-of-the-art gaze prediction model on standard datasets, to obtain an initial gaze prediction map (corresponding to the initial prediction map mentioned above). This map is then weighted and fused with the prior map obtained by the vanishing point detection algorithm to obtain the gaze image for the ATM finding task.

[0070] Specifically, for an input image, an initial gaze prediction map is first extracted using a basic gaze prediction model; simultaneously, a prior map is obtained using a vanishing point detection algorithm, and then prior weights are obtained using a binary classification network. The initial gaze prediction map and the prior map are then weighted and fused to obtain the final gaze prediction map (corresponding to the target prediction map mentioned above). Based on the final gaze prediction map, the placement location of the ATM is selected, thereby improving the rationality of the ATM placement location and increasing the utilization rate of the ATM.

[0071] Example 2

[0072] Embodiment 2 of this application provides an optional device for determining the placement location of an ATM, wherein each implementation unit in the selection device corresponds to each implementation step in Embodiment 1.

[0073] Figure 3 This is a schematic diagram of an optional ATM placement device according to an embodiment of the present invention, such as... Figure 3 As shown, the selection device includes a processing unit 31, a determination unit 32, and a fusion unit 33.

[0074] Specifically, the processing unit 31 is used to input the target image into the target prediction model and output an initial prediction map, wherein the target image is an image of the environment in which the ATM is to be placed, and the initial prediction map includes at least: the gaze point region information of the first type of object predicted by the target prediction model;

[0075] The determining unit 32 is used to calculate the prior map of the target image by using a vanishing point detection algorithm. The prior map includes at least the location of the vanishing point, which is the vanishing point of the target sign in the target image.

[0076] The fusion unit 33 is used to perform weighted fusion of the initial prediction map and the prior map to obtain the target prediction map, wherein the target prediction map is marked with the position of the ATM in the environment to be placed.

[0077] In the ATM placement location determination device provided in Embodiment 2 of this application, the processing unit 31 inputs a target image into a target prediction model and outputs an initial prediction map. The target image is an image of the environment where the ATM is to be placed. The initial prediction map includes at least the gaze point region information of the first type of object predicted by the target prediction model. The determination unit 32 calculates the prior image of the target image using a vanishing point detection algorithm. The prior image includes at least the location of the vanishing point, which is the vanishing point of the target sign in the target image. The fusion unit 33 performs weighted fusion of the initial prediction map and the prior image to obtain a target prediction map, which marks the ATM's placement location in the environment. This solves the technical problem of low ATM utilization caused by manually selecting ATM locations, making it difficult for users to find the ATMs. In this embodiment, an initial prediction map is output through an initial prediction model, and a prior map is determined through a vanishing point detection algorithm. The initial prediction map and the prior map are then weighted and fused to obtain a target prediction map. The placement location of the ATM is selected based on the target prediction map. This avoids the situation in related technologies where the placement location of the ATM is determined by manual selection, which has poor results. Thus, the technical effects of improving the selection efficiency of ATM placement location and improving ATM utilization are achieved.

[0078] Optionally, in the device for determining the placement position of an ATM provided in Embodiment 2 of this application, the determining unit includes: an extraction subunit, used to extract texture features from a target image using a target filter to obtain a first image; a calculation subunit, used to calculate the position of the vanishing point in the first image using a local voting algorithm; and a processing subunit, used to perform Gaussian blur processing on the first image based on the position of the vanishing point and a preset Gaussian weight to obtain a priori image.

[0079] Optionally, in the device for determining the placement position of an ATM provided in Embodiment 2 of this application, the determining unit 32 further includes: preprocessing the target image before extracting texture features from the target image through the target filter, wherein the preprocessing includes at least one of the following: grayscale conversion and noise reduction.

[0080] Optionally, in the device for determining the placement location of an ATM provided in Embodiment 2 of this application, the target prediction model is obtained in the following manner: an acquisition unit is used to acquire an eye-tracking dataset, wherein the eye-tracking dataset includes at least: a target gaze map of a second type of object in the environment, and the target gaze map includes at least: gaze data of the second type of object searching for the ATM in the target image; a training unit is used to train the initial prediction model based on the eye-tracking dataset to obtain the target prediction model.

[0081] Optionally, in the device for determining the placement position of an ATM provided in Embodiment 2 of this application, the acquisition unit includes: a collection subunit for collecting eye movement data of a second type of object searching for an ATM in a target image; a generation subunit for generating a target gaze map based on the eye movement data; and a determination subunit for determining the eye movement dataset based on the target gaze map.

[0082] Optionally, in the device for determining the placement position of an ATM provided in Embodiment 2 of this application, the generation subunit includes: a generation module, used to generate an initial gaze point map of the target image based on eye-tracking data; and a processing module, used to perform a convolution operation on the initial gaze point map using a target Gaussian function to obtain a target gaze point map.

[0083] Optionally, in the device for determining the placement of an ATM provided in Embodiment 2 of this application, the fusion unit 33 includes: an input-output subunit for inputting the target image into the target neural network model and outputting prior weights; and a fusion processing subunit for fusing the initial prediction image and the prior image through the prior weights to obtain the target prediction image.

[0084] Optionally, in the device for determining the placement location of an ATM provided in Embodiment 2 of this application, the target neural network model is obtained in the following manner: a training set acquisition subunit is used to acquire a training set, wherein the training set includes at least: images of a target sign and images of a target sign that does not exist; a training subunit is used to train an initial neural network model using the training set to obtain a target neural network model, wherein the type of the target neural network model includes at least: a binary classification neural network model.

[0085] The aforementioned device for determining the placement of the ATM may also include a processor and a memory. The aforementioned processing unit 31, determining unit 32, and fusion unit 33 are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0086] The aforementioned processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, an initial prediction map is output from the initial prediction model. A vanishing point detection algorithm is used to determine the prior map. The initial prediction map and the prior map are then weighted and fused to obtain the target prediction map. Based on the target prediction map, the placement location of the ATM is selected. This avoids the unsatisfactory results of manually selecting ATM placement locations in related technologies, thus improving the efficiency of ATM placement selection and increasing ATM utilization.

[0087] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0088] According to another aspect of the present invention, an electronic device is also provided, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method for determining the placement position of an ATM as described above by executing the executable instructions.

[0089] According to another aspect of the present invention, a computer-readable storage medium is also provided, which stores a computer program, wherein the computer program controls the device where the computer-readable storage medium is located to execute the method for determining the placement position of an ATM as described above when the computer program is running.

[0090] Figure 4 This is a schematic diagram of an electronic device according to an embodiment of the present invention, such as... Figure 4As shown, an embodiment of the present invention provides an electronic device 40, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the method for determining the placement position of an ATM as described above.

[0091] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0092] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0094] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0096] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0097] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining the placement location of an automatic teller machine, characterized in that, include: The target image is input into the target prediction model, and an initial prediction map is output. The target image is an image of the environment in which the ATM is to be placed. The initial prediction map includes at least the gaze point region information of the first type of object predicted by the target prediction model. The target image is calculated using a vanishing point detection algorithm to determine a priori map of the target image, wherein the priori map includes at least the location of the vanishing point, which is the vanishing point of the target sign in the target image; The initial prediction map and the prior map are weighted and fused to obtain a target prediction map, wherein the target prediction map marks the intended placement location of the ATM in the environment; the target image is input into a target neural network model, which outputs prior weights; the initial prediction map and the prior map are fused using the prior weights to obtain the target prediction map, wherein the prior weights are a measure of the prior tendency of the target image, and the measure is positively correlated with prior information.

2. The determination method according to claim 1, characterized in that, The target image is calculated using a vanishing point detection algorithm to determine its prior image, including: The target image is processed by a target filter to extract texture features, resulting in a first image. The location of the vanishing point in the first image is calculated using a local voting algorithm; Based on the location of the vanishing point and the preset Gaussian weights, the first image is subjected to Gaussian blurring to obtain the prior image.

3. The determination method according to claim 2, characterized in that, Before extracting texture features from the target image using a target filter, the method includes: The target image is preprocessed, wherein the preprocessing includes at least one of the following: grayscale conversion and noise reduction.

4. The determination method according to claim 1, characterized in that, The target prediction model is obtained in the following way: Obtain an eye-tracking dataset, wherein the eye-tracking dataset includes at least: a target gaze map of a second type of object in the environment, the target gaze map including at least: gaze data of the second type of object searching for the ATM in the target image; Based on the eye-tracking dataset, the initial prediction model is trained to obtain the target prediction model.

5. The determination method according to claim 4, characterized in that, Obtaining eye-tracking datasets also includes: The second type of object is used to collect eye-tracking data to locate the ATM in the target image; Based on the eye-tracking data, the target gaze map is generated; The eye-tracking dataset is determined based on the target gaze map.

6. The determination method according to claim 5, characterized in that, Based on the eye-tracking data, the target fixation map is generated, including: Based on the eye-tracking data, an initial fixation map of the target image is generated; The initial gaze map is convolved using a target Gaussian function to obtain the target gaze map.

7. The determination method according to claim 1, characterized in that, The target neural network model is obtained in the following way: Obtain a training set, wherein the training set includes at least: images of the target sign that exist, and images of the target sign that do not exist; The initial neural network model is trained using the training set to obtain the target neural network model, wherein the type of the target neural network model includes at least a binary classification neural network model.

8. A device for determining the placement location of an automatic teller machine, characterized in that, include: The processing unit is used to input the target image into the target prediction model and output an initial prediction map, wherein the target image is an image of the environment in which the ATM is to be placed, and the initial prediction map includes at least: the gaze point region information of the first type of object predicted by the target prediction model; The determining unit is configured to calculate the prior image of the target image by means of a vanishing point detection algorithm, wherein the prior image includes at least the position of the vanishing point, and the vanishing point is the vanishing point of the target sign in the target image; A fusion unit is used to perform weighted fusion of the initial prediction map and the prior map to obtain a target prediction map, wherein the target prediction map marks the placement position of the ATM in the environment; the fusion unit includes: an input-output subunit, used to input the target image into a target neural network model and output prior weights; and a fusion processing subunit, used to perform fusion processing on the initial prediction map and the prior map using the prior weights to obtain the target prediction map, wherein the prior weights are a measure of the prior tendency of the target image, and the measure is positively correlated with prior information.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the method for determining the placement position of the ATM as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The device includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method for determining the placement location of an ATM as described in any one of claims 1 to 7.

Images