Informatization deposit management device and method for tourist attraction

Abstract

The invention discloses an informatization deposit management device and method for a tourist attraction, and relates to the technical field of deposit management. After a left image and a right image are collected through a binocular camera, face area detection is carried out to obtain a left face image and a right face image; obtaining a target disparity map according to the left face image and the right face image, determining a depth change according to the target disparity map, and if the depth change is greater than a preset change, performing feature extraction on the left face image and the right face image to obtain a target feature; and determining to open the corresponding storage cabinet according to the target feature. Face image recognition is carried out on the image collected by the binocular camera to obtain the face area, then the disparity map is determined through the face image to finally obtain the depth change, whether a face recognition object is a real person or a photo can be effectively judged, the safety is improved, and finally the face recognition accuracy is improved by extracting and matching the face depth features. And the corresponding storage cabinet number is accurately found and the unlocking operation is executed, so that the access process is simplified, and the safety is also ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of deposit management, and in particular relates to an information deposit management device and method for tourist attractions. Background Art

[0002] Existing lockers in tourist attractions have many shortcomings in practical applications and are difficult to meet the needs of current tourists and scenic area management. The identity authentication methods of traditional lockers usually rely on simple passwords, credentials or card swiping, which are not safe and reliable in crowded environments of scenic spots. Tourists may face difficulties in retrieving items due to lost credentials or forgotten passwords. Moreover, the password method is easy to be peeped or tampered with by others. Once the credentials are lost or the password is leaked, the deposited items may be taken away by mistake or stolen, and the safety of tourists' personal property cannot be fully guaranteed.

[0003] In order to improve convenience and safety, many scenic spots have tried to introduce facial recognition technology to replace traditional identity authentication methods, using faces for access verification. However, facial recognition also faces new problems in practical applications, especially the risk of photo attacks. Simple facial recognition systems can be easily deceived by static photos, videos or forged images, resulting in accidental unlocking or illegal access. Although facial recognition simplifies the access process to a certain extent, it also brings security risks. Summary of the invention

[0004] The purpose of the present invention is to solve the above-mentioned problems and to provide an information storage management device and method for tourist attractions.

[0005] In the first aspect of the present invention, a method for information storage management for tourist attractions is first proposed, the method comprising:

[0006] When receiving the object-picking command, the binocular camera is turned on to collect images in the target area to obtain left and right images;

[0007] Performing face region detection on the left image and the right image respectively to obtain a first target region and a second target region, cutting the left image according to the first target region to obtain a left face image, and cutting the right image according to the second target region to obtain a right face image;

[0008] Obtaining a target disparity map according to the left face image and the right face image, performing depth change processing on the target disparity map to obtain a target depth image, and extracting a depth change of the target depth image;

[0009] If the depth change is greater than a preset change, extracting features from the left face image and the right face image to obtain target features;

[0010] The preset database is searched according to the target feature to obtain the cabinet number, and the corresponding storage cabinet is opened according to the cabinet number.

[0011] Optionally, performing face area detection on the left image and the right image to obtain the first target area and the second target area includes:

[0012] For an initial face image, preprocessing the initial face image to obtain a preprocessed image, and performing image enhancement on the preprocessed image to obtain a target image;

[0013] Substitute the target image into a preset face detection model to obtain a target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

[0014] Optionally, obtaining a target disparity map according to the left face image and the right face image includes:

[0015] Using a key point detection algorithm, key point recognition is performed on the left face image and the right face image to obtain a first key point set and a second key point set;

[0016] Matching the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtaining a disparity value of the key point combination by a semi-global block matching algorithm;

[0017] The left face image and the right face image are respectively substituted into a real-time stereo matching network to obtain a global disparity map, and the global disparity map is corrected according to the combined disparity values ​​of all key points to obtain a target disparity map.

[0018] Optionally, extracting features from the left face image and the right face image to obtain target features includes:

[0019] For the left face image and the right face image, segmenting the left face image and the right face image through a target grid to obtain a left face block set and a right face block set;

[0020] For each block in the left face block set and the right face block set, extracting features of the block to obtain block features;

[0021] Generate a left face histogram according to all block features in the left face block set, and generate a right face histogram according to all block features in the right face block set;

[0022] The left face histogram and the right face histogram are spliced ​​to obtain a target histogram, and the left face histogram, the right face histogram and the target histogram are substituted into a preset feature fusion model to obtain target features.

[0023] Optionally, the method further includes:

[0024] When receiving a storage instruction, the binocular camera is turned on to collect images in the target area to obtain a first left image and a first right image, and obtain item information;

[0025] Performing face region detection on the first left image and the first right image respectively to obtain a first target acquisition region and a second target acquisition region, cutting the first left image according to the first target acquisition region to obtain a left face acquisition image, and cutting the first right image according to the second target acquisition region to obtain a right face acquisition image;

[0026] Substituting the left face acquisition image and the right face acquisition image into the preset model to obtain target features, and searching for the express cabinet according to the item information to obtain the cabinet number;

[0027] The express cabinet number and the target feature are bound and saved in the preset database.

[0028] In a second aspect of the present invention, an information storage management device for a tourist attraction is provided, comprising:

[0029] The object-picking image acquisition module is used to start the binocular camera to collect images in the target area to obtain left and right images when receiving the object-picking instruction;

[0030] A face image acquisition module, used to perform face region detection on the left image and the right image to obtain a first target region and a second target region, cut the left image according to the first target region to obtain a left face image, and cut the right image according to the second target region to obtain a right face image;

[0031] A depth change determination module, configured to obtain a target disparity map according to the left face image and the right face image, perform depth change processing on the target disparity map to obtain a target depth image, and extract a depth change of the target depth image;

[0032] A target feature determination module, configured to extract features from the left face image and the right face image to obtain target features if the depth change is greater than a preset change;

[0033] The cabinet number search module is used to search a preset database according to the target feature to obtain the cabinet number, and open the corresponding storage cabinet according to the cabinet number.

[0034] Optionally, the face image acquisition module includes:

[0035] An image enhancement module is used to preprocess an initial face image to obtain a preprocessed image, and to enhance the preprocessed image to obtain a target image;

[0036] A target detection frame determination module is used to substitute the target image into a preset face detection model to obtain a target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

[0037] Optionally, the depth change determination module includes:

[0038] A key point recognition module, used to perform key point recognition on the left face image and the right face image respectively by a key point detection algorithm to obtain a first key point set and a second key point set;

[0039] A key point matching module, used for matching the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtaining a disparity value of the key point combination by a semi-global block matching algorithm;

[0040] The target disparity map determination module is used to substitute the left face image and the right face image into the real-time stereo matching network to obtain a global disparity map, and correct the global disparity map according to the combined disparity values ​​of all key points to obtain a target disparity map.

[0041] Optionally, the target feature determination module includes:

[0042] An image segmentation module, for segmenting the left face image and the right face image through a target grid to obtain a left face block set and a right face block set;

[0043] A block feature extraction module, for extracting features of each block in the left face block set and the right face block set to obtain block features;

[0044] A histogram generation module, used to generate a left face histogram according to all block features in the left face block set, and to generate a right face histogram according to all block features in the right face block set;

[0045] The histogram feature fusion module is used to splice the left face histogram and the right face histogram to obtain a target histogram, and substitute the left face histogram, the right face histogram and the target histogram into a preset feature fusion model to obtain target features.

[0046] Optionally, the method further includes:

[0047] The storage image acquisition module is used to start the binocular camera to acquire the image in the target area to obtain the first left image and the first right image, and obtain the item information when receiving the storage instruction;

[0048] A second face image acquisition module, configured to perform face region detection on the first left image and the first right image to obtain a first target acquisition region and a second target acquisition region, cut the first left image according to the first target acquisition region to obtain a left face acquisition image, and cut the first right image according to the second target acquisition region to obtain a right face acquisition image;

[0049] A courier cabinet number generation module, used for substituting the left face acquisition image and the right face acquisition image into the preset model to obtain target features, and searching for the courier cabinet according to the item information to obtain the courier cabinet number;

[0050] The data saving module is used to bind the express cabinet number and the target feature and save them to the preset database.

[0051] Beneficial effects of the present invention:

[0052] The present invention proposes an information storage management method for tourist attractions. When a fetching instruction is received, a binocular camera is turned on to collect images in a target area to obtain a left image and a right image; face area detection is performed on the left image and the right image to obtain a first target area and a second target area, the left image is cut according to the first target area to obtain a left face image, and the right image is cut according to the second target area to obtain a right face image; a target disparity map is obtained according to the left face image and the right face image, depth change processing is performed on the target disparity map to obtain a target depth image, and the depth change of the target depth image is extracted; if the depth change is greater than a preset change, feature extraction is performed on the left face image and the right face image to obtain a target feature; a preset database is searched according to the target feature to obtain a cabinet number, and a corresponding storage cabinet is opened according to the cabinet number. By performing face image recognition on the images collected by the binocular camera to obtain the face area, the data processing volume is reduced and the detection speed is improved. The disparity map is determined through the face image to finally obtain the depth change, which can effectively determine whether the face recognition object is a real person or a photo, thereby improving security. Finally, by extracting and matching the face depth features, the corresponding cabinet number is accurately found and the unlocking operation is performed, which simplifies the access process and ensures safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention will be further described below in conjunction with the accompanying drawings.

[0054] Figure 1 A flowchart of an information-based storage management method for a tourist attraction is provided for an embodiment of the present invention;

[0055] Figure 2 The present invention provides a schematic structural diagram of an information storage management device for a tourist attraction. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the present invention, the description of "first", "second", etc. is only used for descriptive purposes, and cannot be understood as indicating or implying its relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0057] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0058] The embodiment of the present invention provides an information storage management method for tourist attractions. Figure 1 , Figure 1 A flowchart of an information-based storage management method for a tourist attraction provided by an embodiment of the present invention. The method comprises the following steps:

[0059] S101, when receiving a pick-up instruction, the binocular camera is turned on to collect images in the target area to obtain a left image and a right image;

[0060] S102, performing face region detection on the left image and the right image respectively to obtain a first target region and a second target region, cutting the left image according to the first target region to obtain a left face image, and cutting the right image according to the second target region to obtain a right face image;

[0061] S103, obtaining a target disparity map according to the left face image and the right face image, performing depth change processing on the target disparity map to obtain a target depth image, and extracting a depth change of the target depth image;

[0062] S104, if the depth change is greater than the preset change, extracting features from the left face image and the right face image to obtain target features;

[0063] S105, searching a preset database according to the target feature to obtain a storage cabinet number, and opening a corresponding storage cabinet according to the storage cabinet number.

[0064] An information storage management method for tourist attractions provided by an embodiment of the present invention obtains the face area by performing face image recognition on the images collected by the binocular camera, thereby reducing the amount of data processing and improving the detection speed. The disparity map is then determined through the face image to finally obtain the depth change, which can effectively determine whether the face recognition object is a real person or a photo, thereby improving security. Finally, by extracting and matching the face depth features, the corresponding locker number is accurately found and the unlocking operation is performed, thereby simplifying the access process and ensuring security.

[0065] In one implementation, by acquiring images of the target area through a binocular camera, a left-right disparity map can be constructed to extract depth information, which helps to accurately locate the face and reduce the false trigger rate. Facial feature extraction is not started until the depth change reaches a specific threshold, which helps to save computing resources and only performs recognition and matching when the conditions are met, thereby improving the response efficiency of the system.

[0066] In one implementation, the combination of face recognition and depth information can improve recognition accuracy and reduce the possibility of misidentification. Feature extraction and database matching further enhance the reliability of user recognition, thereby reducing the risk of unauthorized access. The target area is the area where the binocular camera can capture faces, and the target area will prompt visitors to stand in the designated position.

[0067] In one implementation, if the depth change is less than or equal to the preset change, it means that the shape of the detection object is relatively flat and there is no significant change in the three-dimensional features. It may not be a real person but a photo. At this time, the subsequent detection processing is stopped.

[0068] In one implementation, the binocular camera will only capture images when it receives an instruction to pick up an item, which reduces resource consumption. Users do not need to perform related operations such as entering a password, simplifying the access process. The preset changes are determined by technical personnel, and the data saved in the preset database is the data generated when the item is stored.

[0069] In one embodiment, performing face region detection on the left image and the right image to obtain the first target region and the second target region includes:

[0070] For the initial face image, preprocessing the initial face image to obtain a preprocessed image, and performing image enhancement on the preprocessed image to obtain a target image;

[0071] Substitute the target image into the preset face detection model to obtain the target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

[0072] In one implementation, image preprocessing and image enhancement can improve image quality, making blurry, poorly lit or poorly contrasted images clear, thereby improving the recognition accuracy of subsequent face detection models, especially under complex lighting or low-resolution conditions; the delineation of the target detection frame enables the face areas in the left and right images to be accurately identified and marked, reducing false detections or missed detections, facilitating disparity analysis in subsequent stereo vision calculations, and ensuring the reliability of depth information calculations.

[0073] In one implementation, after preprocessing and enhancing image quality, the detection model can more effectively identify features, thereby improving detection speed. The optimized process reduces the need for complex models or multiple tests and saves resources.

[0074] In one implementation, preprocessing includes basic format conversion, normalization and other processing; image enhancement uses Enhance-Net to perform low-light enhancement on the left image and the right image respectively, improves the lighting information and details in the image, and generates enhanced left and right images, thereby reducing the loss of details caused by low light and providing a clearer visual basis for subsequent face area detection.

[0075] In one implementation, the preset face detection model may be a face detection model such as Ret i naFace, MTCNN, etc.

[0076] In one embodiment, obtaining a target disparity map according to a left face image and a right face image includes:

[0077] Using a key point detection algorithm, key points of the left face image and the right face image are respectively identified to obtain a first key point set and a second key point set;

[0078] Matching the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtaining a disparity value of the key point combination by a semi-global block matching algorithm;

[0079] The left face image and the right face image are respectively substituted into the real-time stereo matching network to obtain a global disparity map, and the global disparity map is corrected according to the combined disparity values ​​of all key points to obtain a target disparity map.

[0080] In one implementation, the key point matching algorithm can accurately find the corresponding feature points in the left and right images, accurately calibrate the depth information by calculating the disparity value of the key point combination, and effectively reduce errors caused by factors such as lighting and noise; the semi-global block matching algorithm can balance local and global information when calculating the key point disparity, reduce matching errors caused by local occlusion, blur and other problems, and increase the accuracy of stereo matching.

[0081] In one implementation, by using the key point matching results to correct the global disparity map, the accuracy of the disparity map can be greatly improved without losing overall efficiency, while reducing the dependence on complex stereo matching networks; by correcting the global disparity map with the disparity values ​​of the key point combination, the disparity results generated by the real-time stereo matching network can be calibrated, the mismatch area caused by network errors can be reduced, and ultimately a more consistent target disparity map can be obtained.

[0082] In one implementation, the key point detection algorithm can be an Adaboost algorithm based on Haar features, OpenFace, dlib, SURF, etc.; the key points in the first key point set and the second key point set are matched to obtain a key point combination set. Specifically, for each key point, a feature descriptor algorithm such as SI FT, ORB, etc. is used to generate a descriptor, which is used to characterize the local information of the point. The most similar key point pair is found between the key point set of the left image and the key point set of the right image through the KNN matching algorithm, and the Euclidean distance between the two descriptors is calculated. The distance with the smallest Euclidean distance is used as the key point pair obtained by matching.

[0083] In one implementation, the semi-global block matching algorithm may be SGM, Multi-even SGM, Adaptive SGM, Fast SGM, etc.; the real-time stereo matching network may be StereoNet, PSMNet, DeepPruner, Stereo-Unet, etc.

[0084] In one implementation, the global disparity map is corrected according to the disparity values ​​of all key point combinations to obtain the target disparity map. Specifically, the difference in the disparity values ​​of the key point combinations is obtained to obtain a first difference, the difference corresponding to the key point combination in the global disparity map is obtained to obtain a second difference, the difference between the first difference and the second difference is calculated to obtain a correction difference, the pixel distance between the key point combinations is obtained, and the disparity value within each step is corrected with a preset step size as a gradient. For each correction difference of the key point combination, the disparity value is gradually updated according to the preset step size within the pixel distance between them. New, for example, at this time the pixel distance between the key point combination is 100 pixels, the key points are recorded as point A and point B, the preset step size is 20 pixels, and the correction difference is 1. Then, starting from point A to point B, the first 1 to 20 pixels are added with 0.2 on the basis of the formed disparity value, the first 21 to 40 pixels are added with 0.4 on the basis of the formed disparity value, the first 41 to 60 pixels are added with 0.6 on the basis of the formed disparity value, the first 61 to 80 pixels are added with 0.8 on the basis of the formed disparity value, and the first 81 to 100 pixels are added with 1 on the basis of the formed disparity value.

[0085] In one embodiment, extracting features from the left face image and the right face image to obtain target features includes:

[0086] For the left face image and the right face image, the left face image and the right face image are segmented by the target grid to obtain a left face block set and a right face block set;

[0087] For each block in the left face block set and the right face block set, extract features of the block to obtain block features;

[0088] Generate a left face histogram based on all block features in the left face block set, and generate a right face histogram based on all block features in the right face block set;

[0089] The left face histogram and the right face histogram are concatenated to obtain a target histogram, and the left face histogram, the right face histogram and the target histogram are substituted into a preset feature fusion model to obtain the target feature.

[0090] In one implementation, by performing grid segmentation and block feature extraction on the face image, each block represents the local information of the image. This local feature can capture image details such as texture and shape, so that different parts of the face can be more accurately represented. By generating histograms of the left and right faces, and then performing splicing and feature fusion, these local features can be effectively combined to obtain more comprehensive and more recognizable target features. By segmenting the left and right face images into multiple blocks and extracting features from each block, the matching accuracy of facial features can be improved. Traditional overall feature extraction methods may ignore changes in local areas of the image, while the block-based method can perform targeted feature optimization, thereby reducing mismatches.

[0091] In one implementation, the histograms of the left and right faces are spliced ​​to better integrate the visual information from the left and right perspectives. Since the left and right images usually contain different perspective information, feature fusion can combine the advantages of both to improve the comprehensiveness and expression ability of the features. The target features can not only describe the facial features of the individual more accurately, but also integrate the detailed information from different perspectives.

[0092] In one implementation, through target grid segmentation and local feature extraction, the system can adapt to complex environments more flexibly. In some dynamic scenes, different areas of the image may have different backgrounds, lighting or interference factors. By extracting features from local blocks, these interferences can be reduced, improving the ability to recognize objects or faces, thereby improving the stability and speed of recognition.

[0093] In one implementation, the target grid size is determined by a technician, usually a 36 by 36 pixel grid, and feature extraction is performed on the block to obtain the block feature. Specifically, the LBP operator is applied to each image block to extract the local texture features of each small block. The method calculates a binary number by comparing the grayscale value of each pixel with its neighboring pixels, and then converts the binary number to decimal to obtain the corresponding feature value of the pixel point, and obtains the corresponding feature values ​​of all pixels to obtain the block feature.

[0094] In one implementation, since the left and right images are obtained from two different angles, they contain different perspective information. When extracting features, local texture features should be extracted from the left and right images respectively, and then these features are combined to form a complete feature vector. Specifically, the square features of the left and right images are spliced ​​together. For example, the first row of square features in the left image and the first row of square features in the right image are arranged in sequence into a longer square feature, and each row is spliced ​​in turn to obtain the target histogram. This will combine the feature information of the two perspectives and help improve the recognition accuracy, especially under different perspectives and postures.

[0095] In one implementation, the left face histogram, the right face histogram and the target histogram are substituted into a preset feature fusion model to obtain the target feature. Specifically, the left face histogram, the right face histogram and the target histogram are respectively subjected to scale feature extraction at the target scale to obtain a first left face scale map, a first right face scale map and a first target face scale map, wherein the first left face scale map, the first right face scale map and the first target face scale map have the same scale, and the first left face scale map and the first right face scale map are respectively superimposed with the first target face scale map to obtain a first left face fusion map, and the first left face fusion map is obtained. A right face fusion image, and scale feature extraction is performed on the first left face fusion image, the first right face fusion image and the target histogram at the second target scale to obtain a second left face scale image, a second right face scale image and a second target face scale image, and then the third left face scale image and the third right face scale image are obtained according to the same steps above, and this step is repeated until feature extraction is performed at the final target scale to obtain a final left face scale image and a final right face scale image, and the final left face scale image and the final right face scale image are superimposed to obtain a final image, and the target feature is obtained by average pooling the final image through a si gmo id function; the target scale and the final target scale are determined by the technical personnel, and the scale size decreases from the first target scale to the final target scale.

[0096] In one embodiment, the method further comprises:

[0097] When receiving a storage instruction, the binocular camera is turned on to collect images in the target area to obtain a first left image and a first right image, and obtain item information;

[0098] Performing face region detection on the first left image and the first right image respectively to obtain a first target acquisition region and a second target acquisition region, cutting the first left image according to the first target acquisition region to obtain a left face acquisition image, and cutting the first right image according to the second target acquisition region to obtain a right face acquisition image;

[0099] Substitute the left face collection image and the right face collection image into the preset model to obtain the target feature, and find the express cabinet according to the item information to obtain the cabinet number;

[0100] Bind the express cabinet number and target features and save them to the preset database.

[0101] In one implementation, by binding the characteristics of the item with the express cabinet number and combining it with the preset database storage, each item can be tracked and recorded in the system. Managers can easily query the specific location of the item, helping to improve the efficiency of item management. The item information is the information entered by the user, such as the size and weight of the item.

[0102] Based on the same inventive concept, the embodiment of the present invention also provides an information storage management device for tourist attractions. Figure 2 , Figure 2 A schematic diagram of the structure of an information storage management device for a tourist attraction provided by an embodiment of the present invention includes:

[0103] The object-picking image acquisition module is used to start the binocular camera to collect images in the target area to obtain left and right images when receiving the object-picking instruction;

[0104] A face image acquisition module, used to perform face area detection on the left image and the right image respectively to obtain a first target area and a second target area, cut the left image according to the first target area to obtain a left face image, and cut the right image according to the second target area to obtain a right face image;

[0105] A depth change determination module is used to obtain a target disparity map based on the left face image and the right face image, perform depth change processing on the target disparity map to obtain a target depth image, and extract the depth change of the target depth image;

[0106] A target feature determination module, configured to extract features from the left face image and the right face image to obtain target features if the depth change is greater than a preset change;

[0107] The cabinet number search module is used to search the preset database according to the target characteristics to obtain the cabinet number, and open the corresponding storage cabinet according to the cabinet number.

[0108] An information storage management device for tourist attractions provided by an embodiment of the present invention obtains the face area by performing face image recognition on the images collected by the binocular camera, thereby reducing the amount of data processing and improving the detection speed. The disparity map is then determined through the face image to finally obtain the depth change, which can effectively determine whether the face recognition object is a real person or a photo, thereby improving security. Finally, by extracting and matching the face depth features, the corresponding locker number is accurately found and the unlocking operation is performed, thereby simplifying the access process and ensuring security.

[0109] In one embodiment, the face image acquisition module includes:

[0110] An image enhancement module is used for preprocessing the initial face image to obtain a preprocessed image, and performing image enhancement on the preprocessed image to obtain a target image;

[0111] The target detection frame determination module is used to substitute the target image into the preset face detection model to obtain the target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

[0112] In one embodiment, the depth change determination module includes:

[0113] A key point recognition module, used to perform key point recognition on the left face image and the right face image respectively by using a key point detection algorithm to obtain a first key point set and a second key point set;

[0114] A key point matching module is used to match the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtain a disparity value of the key point combination by a semi-global block matching algorithm;

[0115] The target disparity map determination module is used to substitute the left face image and the right face image into the real-time stereo matching network to obtain a global disparity map, and correct the global disparity map according to the combined disparity values ​​of all key points to obtain a target disparity map.

[0116] In one embodiment, the target feature determination module includes:

[0117] An image segmentation module is used to segment the left face image and the right face image through a target grid to obtain a left face block set and a right face block set;

[0118] A block feature extraction module is used to extract features of each block in the left face block set and the right face block set to obtain block features;

[0119] A histogram generation module, used to generate a left face histogram according to all block features in the left face block set, and to generate a right face histogram according to all block features in the right face block set;

[0120] The histogram feature fusion module is used to splice the left face histogram and the right face histogram to obtain a target histogram, and substitute the left face histogram, the right face histogram and the target histogram into a preset feature fusion model to obtain the target feature.

[0121] In one embodiment, the method further comprises:

[0122] The storage image acquisition module is used to start the binocular camera to acquire the image in the target area to obtain the first left image and the first right image, and obtain the item information when receiving the storage instruction;

[0123] A second face image acquisition module is used to perform face area detection on the first left image and the first right image to obtain a first target acquisition area and a second target acquisition area, cut the first left image according to the first target acquisition area to obtain a left face acquisition image, and cut the first right image according to the second target acquisition area to obtain a right face acquisition image;

[0124] The express cabinet number generation module is used to substitute the left face acquisition image and the right face acquisition image into the preset model to obtain the target feature, and to find the express cabinet according to the item information to obtain the express cabinet number;

[0125] The data saving module is used to bind the express cabinet number and target features and save them to the preset database.

[0126] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An information-based storage management method for tourist attractions, characterized in that: The method comprises: When receiving the object-picking command, the binocular camera is turned on to collect images in the target area to obtain left and right images; Performing face region detection on the left image and the right image respectively to obtain a first target region and a second target region, cutting the left image according to the first target region to obtain a left face image, and cutting the right image according to the second target region to obtain a right face image; Obtaining a target disparity map according to the left face image and the right face image, performing depth change processing on the target disparity map to obtain a target depth image, and extracting a depth change of the target depth image; If the depth change is greater than a preset change, extracting features from the left face image and the right face image to obtain target features; The preset database is searched according to the target feature to obtain the cabinet number, and the corresponding storage cabinet is opened according to the cabinet number.

2. The information storage management method for tourist attractions according to claim 1 is characterized in that: Performing face area detection on the left image and the right image to obtain a first target area and a second target area includes: For an initial face image, preprocessing the initial face image to obtain a preprocessed image, and performing image enhancement on the preprocessed image to obtain a target image; Substitute the target image into a preset face detection model to obtain a target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

3. The information storage management method for tourist attractions according to claim 1 is characterized in that: Obtaining a target disparity map according to the left face image and the right face image comprises: Using a key point detection algorithm, key point recognition is performed on the left face image and the right face image to obtain a first key point set and a second key point set; Matching the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtaining a disparity value of the key point combination by a semi-global block matching algorithm; The left face image and the right face image are respectively substituted into a real-time stereo matching network to obtain a global disparity map, and the global disparity map is corrected according to the combined disparity values ​​of all key points to obtain a target disparity map.

4. The information storage management method for tourist attractions according to claim 1 is characterized in that: Extracting features from the left face image and the right face image to obtain target features includes: For the left face image and the right face image, segmenting the left face image and the right face image through a target grid to obtain a left face block set and a right face block set; For each block in the left face block set and the right face block set, extracting features of the block to obtain block features; Generate a left face histogram according to all block features in the left face block set, and generate a right face histogram according to all block features in the right face block set; The left face histogram and the right face histogram are spliced ​​to obtain a target histogram, and the left face histogram, the right face histogram and the target histogram are substituted into a preset feature fusion model to obtain target features.

5. The information storage management method for tourist attractions according to claim 1 is characterized in that: The method further comprises: When receiving a storage instruction, the binocular camera is turned on to collect images in the target area to obtain a first left image and a first right image, and obtain item information; Performing face region detection on the first left image and the first right image respectively to obtain a first target acquisition region and a second target acquisition region, cutting the first left image according to the first target acquisition region to obtain a left face acquisition image, and cutting the first right image according to the second target acquisition region to obtain a right face acquisition image; Substituting the left face acquisition image and the right face acquisition image into the preset model to obtain target features, and searching for the express cabinet according to the item information to obtain the cabinet number; The express cabinet number and the target feature are bound and saved in the preset database.

6. An information storage management device for tourist attractions, characterized in that: The device comprises: The object-picking image acquisition module is used to start the binocular camera to collect images in the target area to obtain left and right images when receiving the object-picking instruction; A face image acquisition module, used to perform face region detection on the left image and the right image to obtain a first target region and a second target region, cut the left image according to the first target region to obtain a left face image, and cut the right image according to the second target region to obtain a right face image; A depth change determination module, configured to obtain a target disparity map according to the left face image and the right face image, perform depth change processing on the target disparity map to obtain a target depth image, and extract a depth change of the target depth image; A target feature determination module, configured to extract features from the left face image and the right face image to obtain target features if the depth change is greater than a preset change; The cabinet number search module is used to search a preset database according to the target feature to obtain the cabinet number, and open the corresponding storage cabinet according to the cabinet number.

7. The information storage management device for tourist attractions according to claim 6 is characterized in that: The face image acquisition module comprises: An image enhancement module is used to preprocess an initial face image to obtain a preprocessed image, and to enhance the preprocessed image to obtain a target image; A target detection frame determination module is used to substitute the target image into a preset face detection model to obtain a target detection frame; if the initial face image is the left image, the target detection frame is the first target area; if the initial face image is the right image, the target detection frame is the second target area.

8. The information storage management device for tourist attractions according to claim 6 is characterized in that: The depth change determination module comprises: A key point recognition module, used to perform key point recognition on the left face image and the right face image respectively by a key point detection algorithm to obtain a first key point set and a second key point set; A key point matching module, used for matching the key points in the first key point set and the second key point set to obtain a key point combination set, and for each key point combination, obtaining a disparity value of the key point combination by a semi-global block matching algorithm; The target disparity map determination module is used to substitute the left face image and the right face image into the real-time stereo matching network to obtain a global disparity map, and correct the global disparity map according to the combined disparity values ​​of all key points to obtain a target disparity map.

9. The information storage management device for tourist attractions according to claim 6, characterized in that: The target feature determination module comprises: An image segmentation module, for segmenting the left face image and the right face image through a target grid to obtain a left face block set and a right face block set; A block feature extraction module, for extracting features of each block in the left face block set and the right face block set to obtain block features; A histogram generation module, used to generate a left face histogram according to all block features in the left face block set, and to generate a right face histogram according to all block features in the right face block set; The histogram feature fusion module is used to splice the left face histogram and the right face histogram to obtain a target histogram, and substitute the left face histogram, the right face histogram and the target histogram into a preset feature fusion model to obtain target features.

10. The information storage management device for tourist attractions according to claim 6, characterized in that: The method further comprises: The storage image acquisition module is used to start the binocular camera to acquire the image in the target area to obtain the first left image and the first right image, and obtain the item information when receiving the storage instruction; A second face image acquisition module, configured to perform face region detection on the first left image and the first right image to obtain a first target acquisition region and a second target acquisition region, cut the first left image according to the first target acquisition region to obtain a left face acquisition image, and cut the first right image according to the second target acquisition region to obtain a right face acquisition image; A courier cabinet number generation module, used for substituting the left face acquisition image and the right face acquisition image into the preset model to obtain target features, and searching for the courier cabinet according to the item information to obtain the courier cabinet number; The data saving module is used to bind the express cabinet number and the target feature and save them to the preset database.

Images