Lost and found method and system based on intelligent bus platform

Abstract

The application relates to the technical field of wisdom platforms, in particular to a lost-and-found method and system based on a wisdom bus platform, which comprises the following steps: obtaining target lost object information of a lost object, analyzing the target lost object, and generating a stationary object set; generating a semantic image description based on the stationary object set, and generating a candidate lost object description set based on a preset semantic description generation function; performing description similarity matching based on the candidate lost object description set to obtain a suspected target set; performing trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model to obtain a lost object judgment result. The application structures the spatial trajectory of a candidate object in a carriage, the stationary state change and the passenger riding behavior trajectory by using the trajectory matching scoring model, clearly defines the corresponding relationship between the passenger riding behavior and the object trajectory, improves the identification accuracy of lost object identification, and improves the accuracy of the judgment result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wisdom platform, in particular to a lost and found method and system based on a wisdom bus platform. BACKGROUND

[0002] With the continuous development of intelligent transportation systems, the bus platform gradually has the ability of multi-source data fusion such as vehicle video collection, ticket information management and in-car positioning. On this basis, for the automatic identification and recall of passenger lost objects, the market has tried to build a lost and found function module based on image recognition, semantic tags and time period analysis to help passengers find lost objects.

[0003] The existing lost object identification has deficiencies, specifically, the existing lost object identification method mostly compares and identifies lost objects according to images or passenger descriptions. For example, an invention patent application with publication number CN114863516A published on August 5, 2022 discloses a public transportation lost article retrieval method and system based on image processing. First image information of passengers entering public transportation is obtained, and based on the first image information, boarding data tuples are extracted. Second image information of passengers leaving public transportation is obtained, and based on the second image information, alighting data tuples are extracted. The boarding data tuples and the alighting data tuples each include identity recognition feature data and carrying article feature data of the corresponding passenger. Then, the alighting data tuples corresponding to the boarding data tuples of the passengers are obtained based on the identity recognition feature data, and the carrying article feature data in the boarding data tuples and the alighting data tuples are compared. If the carrying article feature data in the boarding data tuples has no corresponding matching item in the alighting data tuples, it is determined that the passenger has lost an article when leaving the public transportation, and a reminder is given.

[0004] Although it can identify the features reduced by the passengers when getting off the bus by comparing the images of the passengers when getting on and off the bus, it can remind the passengers that they may have lost an article when they get off the bus, and improve the probability of finding the article, but it and other solutions in the prior art have similar problems, specifically, the corresponding relationship between the passenger's riding behavior and the object's trajectory is ignored, resulting in a large amount of redundancy of candidate targets, and ultimately leading to low accuracy of the determination result, thereby further leading to low accuracy of the lost and found. SUMMARY

[0005] Therefore, it is necessary to provide a lost and found method and system based on a wisdom bus platform, which can structure the spatial trajectory of the candidate object in the car, the change of the stationary state and the trajectory of the passenger's riding behavior by a trajectory matching scoring model, clearly define the corresponding relationship between the passenger's riding behavior and the object's trajectory, improve the identification accuracy of the lost object identification, and improve the accuracy of the determination result.

[0006] The technical scheme of the present application is as follows:

[0007] A lost and found method based on a smart bus platform, the method comprising:

[0008] Obtaining target lost object information of a lost object, analyzing the target lost object, and generating a stationary object set;

[0009] Generating semantic image description based on the stationary object set, and generating a candidate lost object description set based on a preset semantic description generation function;

[0010] Performing description similarity matching based on the candidate lost object description set to obtain a suspected target set;

[0011] Performing trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model to obtain a lost object judgment result.

[0012] Specifically, obtaining target lost object information of a lost object, analyzing the target lost object, and generating a stationary object set, comprises:

[0013] Obtaining target lost object information of a lost object, and generating a target segment set based on the target lost object information;

[0014] Performing image feature analysis based on the target segment set to obtain a stationary object set.

[0015] Specifically, performing image feature analysis based on the target segment set to obtain a stationary object set, comprises:

[0016] Extracting a spatial feature vector based on the target segment set, and calculating the spatial similarity of the spatial feature vectors between frame images based on the spatial feature vector;

[0017] Obtaining an image matrix based on the target segment set, and generating a time stability score based on the image matrix;

[0018] Generating a space-time stability score of the lost object based on the spatial similarity and the time stability score;

[0019] Obtaining a stationary object set based on the space-time stability score.

[0020] Specifically, generating semantic image description based on the stationary object set, and generating a candidate lost object description set based on a preset semantic description generation function, comprises:

[0021] Extracting an image feature vector of a stationary object based on the target segment set;

[0022] Generating a matching weight of an anchor point based on the two-dimensional position coordinates of the stationary object, wherein the anchor point is pre-configured;

[0023] setting a semantic label of an anchor point with the largest weight as a semantic reference of the stationary object;

[0024] generating a sentence description based on the image feature vector and the semantic label of the anchor point according to a semantic description generation function,

[0025] generating a candidate lost property description set according to the sentence description.

[0026] Specifically, description similarity matching is performed according to the candidate lost property description set to obtain a suspected target set, including:

[0027] obtaining description information provided by the passenger at the time of the lost property declaration request;

[0028] performing description similarity matching on the candidate lost property description set and the description information provided at the time of the declaration based on a preset semantic encoder to obtain a suspected target set.

[0029] Specifically, trajectory consistency analysis is performed on the suspected target set based on a preset trajectory matching scoring model to obtain a lost property judgment result, including:

[0030] generating an object continuous stationary probability, a stationarity score and a spatial overlap degree according to the suspected target set;

[0031] generating a trajectory consistency score based on the object continuous stationary probability, the stationarity score and the spatial overlap degree according to a preset trajectory matching scoring model;

[0032] obtaining a lost property judgment result according to the trajectory consistency score.

[0033] Specifically, a lost property judgment result is obtained according to the trajectory consistency score, including:

[0034] determining whether the trajectory consistency score is greater than a preset high-confidence threshold;

[0035] If the determination is yes, the candidate object is determined to be a high-confidence lost property, and the high-confidence lost properties are aggregated to generate a lost property judgment result.

[0036] Specifically, a lost property finding system based on a smart public transportation platform is also provided, and the system includes:

[0037] a stationary object generation module configured to obtain target lost property information of a lost item, analyze the target lost property, and generate a stationary object set;

[0038] a lost property description generation module configured to generate semantic image description based on the stationary object set, and generate a candidate lost property description set based on a preset semantic description generation function;

[0039] a suspect target generation module configured to perform description similarity matching on the candidate lost property description set to obtain a suspect target set;

[0040] a judgment result generation module configured to perform trajectory consistency analysis on the suspect target set based on a preset trajectory matching scoring model to obtain a lost property judgment result.

[0041] Specifically, the stationary object generation module is further configured to: obtain target lost property information of a lost item, and generate a target segment set according to the target lost property information; perform image feature analysis on the target segment set to obtain a stationary object set.

[0042] Specifically, the stationary object generation module is further configured to: extract a spatial feature vector from the target segment set, calculate a spatial similarity of the spatial feature vectors between frame images according to the spatial feature vector, obtain an image matrix from the target segment set, and generate a time stability score according to the image matrix; generate a spatiotemporal stability score of the lost object according to the spatial similarity and the time stability score; and obtain the stationary object set according to the spatiotemporal stability score.

[0043] Specifically, the lost property description generation module is further configured to: extract an image feature vector of the stationary object from the target segment set; generate a matching weight of an anchor point according to a two-dimensional position coordinate of the stationary object, wherein the anchor point is preconfigured; set a semantic label of an anchor point with the largest weight as a semantic reference of the stationary object; generate a sentence description according to the image feature vector and the semantic label of the anchor point based on a semantic description generation function; and generate a candidate lost property description set according to the sentence description.

[0044] Specifically, the suspect target generation module is further configured to: obtain description information provided by a passenger at a lost property declaration request; and perform description similarity matching on the candidate lost property description set and the description information provided at the lost property declaration request based on a preset semantic encoder to obtain a suspect target set.

[0045] Specifically, the judgment result generation module is further configured to: generate an object continuous stationary probability, a stationarity score, and a spatial overlap degree according to the suspect target set; generate a trajectory consistency score based on a preset trajectory matching scoring model according to the object continuous stationary probability, the stationarity score, and the spatial overlap degree; and obtain a lost property judgment result according to the trajectory consistency score.

[0046] Specifically, the judgment result generation module is further configured to: determine whether the trajectory consistency score is greater than a preset high-confidence threshold; if the determination is yes, determine that a candidate object is a high-confidence lost property; and generate a lost property judgment result by aggregating each high-confidence lost property.

[0047] Optionally, the application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the lost and found method based on the smart bus platform when executing the computer program.

[0048] Optionally, the application further provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the lost and found method based on the smart bus platform when executed by a processor.

[0049] The application relates to machine learning and deep learning technology, and achieves the following technical effects:

[0050] The target lost object information of the lost object is acquired in sequence, the target lost object is analyzed, and a stationary object set is generated; semantic image description generation is performed according to the stationary object set, and a candidate lost object description set is generated based on a preset semantic description generation function; description similarity matching is performed according to the candidate lost object description set, and a suspected target set is obtained; trajectory consistency analysis is performed on the suspected target set based on a preset trajectory matching scoring model, and a lost object judgment result is obtained; and then the trajectory matching scoring model is used to structurally model the spatial trajectory of the candidate object in the carriage, the stationary state change and the passenger riding behavior trajectory, the corresponding relationship between the passenger riding behavior and the object trajectory is determined, the spatiotemporal correlation between the object and the passenger is judged through a scoring mechanism, redundant data is reduced, the identification accuracy of the lost object identification is improved, and the accuracy of the judgment result is improved. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 It is a flowchart of the lost and found method based on the smart bus platform in one embodiment;

[0052] Figure 2 It is a structural block diagram of the lost and found system based on the smart bus platform in one embodiment. DETAILED DESCRIPTION

[0053] In the following description, specific details are set forth in order to provide a thorough understanding of the application. However, persons having ordinary skill in the art will appreciate that embodiments of the application can be practiced without the specific details, and that the application is not limited to particular embodiments described.

[0054] It should be understood that the word “comprise” or variations such as “comprises” or “comprising”, when used in this specification and in the accompanying claims, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0055] It should also be understood that the term “and / or” when used in this specification and in the following claims is intended to mean one or the other or both of the associated listed items and that no combinations of one or more items included in the processes, methods, systems or computer readable media are intended to be excluded.

[0056] As used in this specification and claims, the terms “if’ and “when” can be interpreted to mean “upon” or “in response to a determination” or “in response to a detection” depending on the context. Similarly, the phrase “if it is determined” or “if [a described condition or event] is detected” can be interpreted to mean “upon determining” or “in response to a determining” or “upon detecting [the described condition or event]” or “in response to a detection [of the described condition or event]” depending on the context.

[0057] In addition, the terms “first”, “second”, “third”, etc. are used in the description and claims of this application merely to distinguish one element from another and are not intended to imply or create any relative importance of the elements.

[0058] Reference throughout this specification to “one embodiment” or “some embodiments” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearances of the phrases “in one embodiment”, “in some embodiments”, “in other embodiments”, “in additional embodiments”, and so on, in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specifically stated. The terms “comprise”, “comprises”, “comprising”, “include”, “includes”, “including” and “contain”, “contains”, “containing” and variations thereof are inclusive and are meant to be equivalent to the term “comprising” unless otherwise specifically stated.

[0059] In one embodiment, a terminal is provided, the terminal is configured to: acquire target lost article information of a lost article, analyze the target lost article, and generate a stationary object set; perform semantic image description generation according to the stationary object set, and generate a candidate lost article description set based on a preset semantic description generation function; perform description similarity matching according to the candidate lost article description set, to obtain a suspected target set; perform trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model, to obtain a lost article judgment result.

[0060] The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices.

[0061] In one embodiment, as shown in Figure 1 A lost and found method based on a smart bus platform is provided, and the method comprises the following steps:

[0062] Step S100: target lost object information of a lost object is acquired, the target lost object is analyzed, and a stationary object set is generated;

[0063] Step S200: semantic image description generation is performed according to the stationary object set, and a candidate lost object description set is generated based on a preset semantic description generation function;

[0064] Step S300: description similarity matching is performed according to the candidate lost object description set, and a suspected target set is obtained;

[0065] Step S400: trajectory consistency analysis is performed on the suspected target set based on a preset trajectory matching scoring model, and a lost object judgment result is obtained.

[0066] The present application acquires target lost object information of a lost object, analyzes the target lost object, and generates a stationary object set; performs semantic image description generation according to the stationary object set, and generates a candidate lost object description set based on a preset semantic description generation function; performs description similarity matching according to the candidate lost object description set, and obtains a suspected target set; performs trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model, and obtains a lost object judgment result; and further realizes structured modeling of spatial trajectories, stationary state changes and passenger riding behavior trajectories of candidate objects in a car compartment through a trajectory matching scoring model, clearly defines the corresponding relationship between passenger riding behavior and object trajectories, judges the spatiotemporal correlation between objects and passengers through a scoring mechanism, reduces redundant data, improves the recognition accuracy of lost object recognition, and improves the accuracy of the judgment result.

[0067] In one embodiment, step S100: target lost object information of a lost object is acquired, the target lost object is analyzed, and a stationary object set is generated, comprising:

[0068] Step S110: target lost object information of a lost object is acquired, and a target segment set is generated according to the target lost object information;

[0069] Step S120: image feature analysis is performed according to the target segment set, and a stationary object set is obtained.

[0070] In this embodiment, in order to clearly define the type of lost property and accurately limit the spatiotemporal range where the lost property may appear, accurate boundary conditions and screening basis are provided for the subsequent video frame segment extraction, image feature analysis and trajectory consistency analysis steps. Therefore, the target lost property information of the lost property is obtained, and the target segment set is generated according to the target lost property information.

[0071] After the lost property, the passenger inputs the description of the target lost property related information through the intelligent bus platform, which includes the type of lost property, boarding time, alighting time, bus line number, boarding and alighting station, and the position of the lost property. The type of lost property is the description of the passenger on the type and state of the lost property, for example, a large black backpack containing something, and the position of the lost property is described as the middle of the vehicle compartment and the rear seat.

[0072] Based on the input bus line number, boarding time, alighting time and boarding and alighting station, the platform queries the vehicle running on the line in the corresponding time period to determine the unique corresponding vehicle number, ensuring the uniqueness and accuracy of the video data extraction in the subsequent steps. After confirming the vehicle number, the platform retrieves the running trajectory data of the vehicle, uses the GPS trajectory playback technology in the prior art to obtain the complete driving path of the vehicle in the time period from the boarding time to the alighting time, confirms the stop station of the vehicle, and the driving path data includes the time stamp, GPS coordinates, speed and stop state of the corresponding time every second. Based on the number and installation position of the cameras in the vehicle, a vehicle compartment space structure diagram is established, the position of the lost property described by the passenger is mapped to one or more specific camera view areas, the set of camera numbers corresponding to the position of the lost property is confirmed, and finally the target lost property information is obtained, including the vehicle number, the complete driving path data in the time period from the boarding time to the alighting time, and the set of camera numbers corresponding to the position of the lost property.

[0073] Then, according to the target lost article information, a vehicle-mounted video frame segment extraction is performed to obtain a target segment set. In the vehicle-mounted video frame segment extraction, according to the target lost article information, a vehicle number and a frame index range of the vehicle-mounted monitoring video in a corresponding time period are obtained, and a video frame segment corresponding to the time period from the boarding time to the alighting time and the camera number is accurately extracted from the vehicle-mounted monitoring video stored in the multi-channel video storage system of the vehicle. Specifically, by using a multi-channel time synchronization index retrieval technology based on the vehicle-mounted DVR system, the technology has high stability and accurate time alignment capability. First, multi-channel camera recognition and channel mapping are performed. The video acquisition system inside the bus vehicle is composed of multiple fixed cameras, which are often connected to the vehicle-mounted DVR device to form a channel index table. Each camera channel corresponds to a fixed installation point, and the number is one-to-one corresponding to the carriage space model. According to the obtained camera number, the video extraction range is limited to the relevant view angle of the passenger description, avoiding irrelevant data redundancy processing, improving the efficiency and reducing the subsequent identification difficulty; then, time segment extraction and frame index construction are performed. The vehicle-mounted DVR system is attached with high-precision time stamps for each video, and the system maps the corresponding time period to the frame index interval of each channel. With the aid of the synchronous index structure of the DVR, the system can achieve millisecond-level alignment, ensuring that the extracted video frame segment completely covers the window in which the lost article may appear in time. Then, data segmentation and frame structure generation are performed. The system extracts continuous frame images from each camera channel according to a fixed frame rate (such as 25 fps) to form a frame sequence in the corresponding time period, and adds the camera number, the capture time of each frame, etc. In addition, the inter-frame average brightness for abnormal occlusion detection and the inter-frame motion estimation for auxiliary judgment of the stationary state can also be added. Finally, the target segment set is output, including the camera number, the corresponding time period and the frame sequence in the corresponding time period, providing a high-precision and low-redundancy video basis for image feature extraction. The target segment set includes a plurality of target segments, each target segment includes a plurality of images, each target segment has a corresponding time period, and each image corresponds to a corresponding frame of the frame sequence in the corresponding time period, i.e. each frame in the corresponding time period of each target segment corresponds to an image. Then, according to the target segment set, image feature analysis is performed to obtain a stationary object set.

[0074] In one embodiment, step S120: according to the target segment set, image feature analysis is performed to obtain a stationary object set, including:

[0075] Step S121: according to the target segment set, spatial feature vector extraction is performed, and the spatial similarity of the spatial feature vectors between the frame images is calculated according to the spatial feature vectors;

[0076] Step S122: according to the target segment set, an image matrix is obtained, and a time stability score is generated according to the image matrix;

[0077] Step S123: generating a spatio-temporal stability score of the missing object according to the spatial similarity and the temporal stability score;

[0078] Step S124: obtaining a stationary object set according to the spatio-temporal stability score.

[0079] In this embodiment, the image feature analysis includes: using a spatio-temporal stability analysis model to perform image feature analysis according to the target segment set, identifying a stationary object, and obtaining a stationary object set. The stationary object set contains relevant information of all stationary objects, including image region feature vectors of the object, frame position information, position coordinate information, camera number, etc.

[0080] Specifically, the spatial feature vector of each frame image of the target segment set is extracted. The color, texture, shape, etc. features are extracted through a convolutional neural network (CNN) model, and the spatial similarity of the spatial feature vectors between the frame images is calculated. The convolutional neural network (CNN) model can use ResNet-50, use fixed parameters, and does not need to be trained to ensure stability and reproducibility.

[0081] The calculation of the spatial similarity is as follows:

[0082]

[0083] is the spatial similarity between the spatial feature vectors extracted from the kth frame and the k+1th frame images. The higher the similarity, the more stable the object is in the spatial position, and it is possible to be a stationary object. is the spatial feature vector extracted from the kth frame image, is the spatial feature vector extracted from the k+1th frame image. is the Euclidean distance, i.e., the L2 norm.

[0084] Next, based on a time stability function, an image matrix is obtained according to the target segment set, and a time stability score is generated according to the image matrix. The time stability function is as follows:

[0085]

[0086] wherein, is the time stability score, the higher the score, the more stable the object is in the time period, and it is possible to be a stationary object, is the maximum step length of the front and back frame difference calculation, is the image matrix of the i-th segment, the j-th camera, and the k+n-th frame, which represents the difference with the k-th frame image matrix in the time dimension, is the image matrix of the i-th segment, j-th camera, k-th frame, is the Euclidean distance of the gray value difference of all pixels between the k-th frame and k+n-th frame image, which is obtained by squaring, summing and square root of the gray value difference of each pixel, reflecting the overall brightness difference between the two frames of images, is the Euclidean distance, i.e. L2 norm.

[0087] Based on the spatiotemporal stability analysis model, the spatiotemporal stability score of the missing object is generated according to the spatial similarity and the temporal stability score. The spatiotemporal stability analysis model is as follows:

[0088]

[0089] wherein, is the spatiotemporal stability score of the object, which is used to determine whether it is a stationary object, is a temporal stability function, is the image matrix of the i-th segment, j-th camera, k-th frame, is the spatial feature vector extracted from the k-th frame image, is the spatial feature vector extracted from the k+1-th frame image, is a spatial similarity function between spatial feature vectors. is a weight coefficient for balancing the temporal stability and the spatial similarity, the range of the weight coefficient is [0, 1], and the weight coefficient is set according to whether there is occlusion, light interference and whether the image is clear and stable. If there is occlusion or there is a window light reflection, the weight is reduced (such as 0.3 to 0.5), and the sensitivity to brightness is weakened. If the picture clarity is poor, it is difficult to extract spatial features, and the weight is increased (such as 0.6 to 0.8), and the influence of the spatial similarity is reduced.

[0090] In the prior art, a single spatial feature such as color, texture, shape, etc. is usually relied on to determine the object type, and the temporal information in the video is ignored, especially the key factor of whether the object is stationary. In the present embodiment, the stationary nature of the object in time and the spatial features are considered at the same time, the spatial features of each frame of image and the temporal stability of the object, i.e. whether the object remains stationary in multiple frames, are combined, and multi-dimensional feature fusion of the image is performed, thereby improving the detection accuracy of the stationary object and improving the accuracy and robustness of object recognition.

[0091] In one embodiment, step S200: generating a semantic image description according to the stationary object set, and generating a candidate lost property description set based on a preset semantic description generation function, comprising:

[0092] Step S210: generating an image feature vector of the stationary object according to the target segment set for image feature extraction;

[0093] Step S220: generating a matching weight of an anchor point according to a two-dimensional position coordinate of the stationary object, wherein the anchor point is pre-configured;

[0094] Step S230: setting a semantic label of an anchor point with the largest weight as a semantic reference of the stationary object;

[0095] Step S240: generating a sentence description according to an image feature vector and a semantic label of an anchor point based on a semantic description generation function,

[0096] Step S250: generating a candidate lost property description set according to the sentence description.

[0097] In the embodiment, when generating a semantic image description, first, a vehicle space structure anchor point is constructed, a semantic label capable of being described is added to the space structure anchor point in the vehicle, each anchor point contains a two-dimensional position coordinate of the anchor point and a semantic label of the anchor point (for example, “front door”, “seat right side”, “vehicle tail ground”), then using an anchor point matching weight function, the stationary object is associated and matched with the nearest anchor point, so that the semantic label of the anchor point is used as the semantic reference of the stationary object, finally, using a semantic description generation function, the image features of the stationary object are jointly coded with the semantic label of the anchor point to generate a semantic image description, all semantic image descriptions are combined to obtain a candidate lost property description set.

[0098] Specifically, when constructing the vehicle space structure anchor point, the vehicle space structure anchor point set is wherein, is a space structure anchor point in the set, containing a two-dimensional position coordinate of the anchor point and a semantic label information of the anchor point. Specifically, ,

[0099] wherein, is a two-dimensional position coordinate of the anchor point j, that is, a position of a center point of the object in an image coordinate system, is a semantic label of the anchor point j, such as “front door”, “seat right side”, “vehicle tail ground”, used for describing a position of the anchor point in the vehicle space structure. j is an index of the space structure anchor point.

[0100] Further, according to the target segment set, an image feature of the stationary object is extracted to generate an image feature vector of the stationary object, after obtaining the target segment set, a convolutional neural network (CNN) model such as ResNet-50 is used to extract image features of each frame image in the segment set to obtain an image feature vector of the stationary object Then, a matching weight of an anchor point is generated according to a two-dimensional position coordinate of the stationary object wherein the anchor point is pre-configured. The matching weight is generated based on the following anchor point matching weight function:

[0101]

[0102] wherein, is the matching weight of the stationary object and the anchor point, is the two-dimensional position coordinate of the stationary object, is the square of the Euclidean distance between the object and the anchor point, the smaller the value, the shorter the distance between the object and the anchor point, is the distance decay control factor. The distance decay control factor is used to control the distance sensitivity when matching the stationary object and the anchor point. It is usually a fixed value set according to the expected spatial tolerance range (in pixels) near the anchor point. For example, for a high-definition image, a larger spatial drift can be tolerated, so the spatial tolerance range is set to 60, and the distance decay control factor is 1 / 60 2 , and in small target detection, in order to improve the accuracy, the spatial tolerance range is set to 20, and the distance decay control factor is 1 / 20 2 , by setting a negative sign in the exp function, the stationary object is associated with the nearest anchor point, and the anchor point weight is larger when the distance is closer. Then set the semantic label of the anchor point with the largest weight as the semantic reference of the stationary object; the sentence description is generated based on the semantic description generation function according to the image feature vector and the semantic label of the anchor point .

[0103] wherein the sentence description is generated based on the semantic description generation function :

[0104]

[0105]

[0106] wherein, is the sentence description, is the image feature vector of the stationary object, is the semantic label of the anchor point with the largest anchor point matching weight, is the image feature mapping function, which is used to convert the image features of the stationary object into a semantic embedding vector, and is realized by linear transformation plus nonlinear activation (such as ReLU), is the anchor label embedding function, which is used to convert the semantic label of the anchor point into an embedding vector, and is realized by word vector embedding (such as Word2Vec), is a vector splicing operation, is a language generation decoder used to generate the final description text.

[0107] Therefore, by converting the image features and the semantic label of the anchor point After embedding, splicing, and inputting to the language generation decoder Generating sentence descriptions For example, "a blue handbag is placed under the 4th row of seats", generate a candidate lost property description set for each stationary object.

[0108] In this embodiment, the semantic description of the stationary object is generated by matching the association between the stationary object and the nearest anchor point, combined with the image features of the stationary object. The existing image description generation methods such as CNN+LSTM or Transformer-based image caption generation model usually cannot understand the structural spatial position of the carriage (such as "next to the front door" and "under the seat") in a specific closed environment such as a carriage, it is difficult to generate practical descriptions useful to passengers, and can only generate descriptions such as "a bag", lacking spatial adaptability and practicality. In this application, a set of spatial structure anchor points is introduced, which has the ability to perceive the position, adapt to the closed space, improve the spatial adaptability, and generate semantic image descriptions, generate descriptions useful to passengers, such as "a black backpack near the back door", and improve the practicality.

[0109] In one embodiment, step S300: performing description similarity matching according to the candidate lost property description set to obtain a suspected target set, comprising:

[0110] Step S310: obtaining the description information provided by the passenger at the time of lost property declaration request;

[0111] Step S320: performing description similarity matching on the candidate lost property description set and the description information at the time of declaration based on a preset semantic encoder to obtain a suspected target set.

[0112] In this embodiment, each description in the candidate lost property description set corresponds to an identified stationary object. The description information provided by the passenger when initiating the lost property report request on the smart bus platform is matched with the description through a semantic encoder SBERT (Sentence-BERT) combined with cosine similarity calculation. According to the similarity matching result, the descriptions of all stationary objects that meet the conditions, i.e., the descriptions of stationary objects with a similarity value greater than a set threshold, are added to the suspected target set. If necessary, the maximum number of additions can be set, and the additions are sorted from high to low similarity and added to the suspected target set to control the consumption of computing resources. The set threshold is set according to experience, for example, a similarity greater than 0.75 can be considered a suspected target. The semantic encoder SBERT (Sentence-BERT) is a mature and stable natural language semantic representation method with good generalization ability and efficient online computing characteristics. By encoding the passenger description and the lost property description, a semantic vector is obtained, and the cosine similarity is calculated to obtain the similarity between the two descriptions. This method can capture contextual semantics, support complex description expression, adapt to different passenger language expression methods, has fixed model parameters, and can run without complex tuning, with good generalization ability and efficient online computing characteristics.

[0113] In one embodiment, step S400: performing trajectory consistency analysis on the suspected target set based on a preset trajectory matching score model to obtain a lost property judgment result, including:

[0114] Step S410: generating object continuous stationary probability, stationarity score, and spatial overlap degree based on the suspected target set;

[0115] Step S420: generating a trajectory consistency score based on the object continuous stationary probability, stationarity score, and spatial overlap degree based on a preset trajectory matching score model;

[0116] Step S430: obtaining a lost property judgment result based on the trajectory consistency score.

[0117] In this embodiment, each element in the suspected target set corresponds to a candidate identified as a stationary object. Since the lost property may experience multiple stationary, moving, and occluded states during the operation of the bus, it is difficult to ensure accuracy relying solely on static images and semantic matching. Therefore, the time series trajectory of the object in the vehicle cabin video and the passenger trajectory are combined through a trajectory matching score model to perform trajectory consistency analysis on each target object in the set to obtain a judgment result and perform lost property claim based on the judgment result.

[0118] Specifically, the object continuous stationary probability , stationarity score and spatial overlap . Wherein, represents the object stationary probability in the object loss time period, the greater the value, the more stationary the object is in the time period, reflecting the possibility that the lost object is not taken away, moved, and can also be used to exclude cases such as passengers carrying objects mistaken for lost objects.

[0119] stationary degree score represents the stationary degree score of the object trajectory in the object loss time period, the greater the value, the higher the stationary degree score, which can effectively exclude non-stationary targets such as objects placed by humans and then picked up.

[0120] spatial overlap . Represents the spatial overlap of the passenger trajectory area and the object appearance position, the greater the value, the more likely the passenger is to lose the object near the object position.

[0121] wherein the object stationary probability is generated comprises:

[0122] ;

[0123] wherein, is the object stationary probability, is the spatial trajectory sequence of the object, including the timestamp and two-dimensional position coordinates; is the mth frame image in the object loss time period; is the object loss time period described by the passenger; is the mode of the object position (i.e. the center coordinates) in the object loss time period, i.e. the most common position; is the fluctuation of the object position in the object loss time period, if the distance difference between the spatial position of the object in each frame image in the time period and the mode of the object position is less than the set threshold, i.e. the position fluctuation tolerance range (such as 5 pixels), then the value is 1, otherwise the value is 0; is the total number of frame images in the object loss time period. represents that the center coordinates of the object in the mth frame image in the object loss time period belong to , the mth frame image is in this time period.

[0124] generate the stationary degree score as follows:

[0125]

[0126] wherein the stationary degree control factor is used to normalize the square of the calculated Euclidean distance between objects to a dimensionless number, which is set according to the tolerance degree of the trajectory fluctuation. , is the tolerance of trajectory fluctuation, the higher the tolerance of trajectory fluctuation, the greater the tolerance, the higher the smoothness score under the same trajectory fluctuation, for example, the tolerance of trajectory fluctuation is higher, the tolerance is set to 20 pixels, the tolerance of trajectory fluctuation is very sensitive, the tolerance is low, and the tolerance is set to 5 pixels; : two-dimensional position coordinates of the object in the previous frame image of the object; : the square of the Euclidean distance of the object between two frame images, reflecting the movement of the object.

[0127] represents the square value of the average inter-frame movement distance of the object in the time period, if the object is stationary, the value is equal to or close to 0, if the object is constantly moving, the value changes significantly. The square of the Euclidean distance is a geometric standard of two-dimensional translation, which is direction-independent and convenient for subsequent processing. Squaring avoids positive and negative offset, and more reasonably reflects the movement intensity.

[0128] The spatial overlap degree includes calculating the overlap ratio of the passenger riding area (inferred by ticket information, boarding and alighting point) and the object appearance area, and the specific calculation formula is as follows:

[0129] ;

[0130] wherein, is the number of regions where the object appearance area and the passenger possible activity area overlap (the car space can be divided into 20*20 grids), which is obtained according to the number of regions where the object is captured in the passenger possible activity area; is the number of passenger possible activity areas, since each car compartment of the bus is divided into multiple camera monitoring blocks, the number of passenger possible activity areas can be found according to the number of times the passenger is captured by different cameras during the passenger's riding time period.

[0131] Finally, based on the preset trajectory matching score model, the trajectory consistency score is generated according to the object continuous stationary probability, the smoothness score and the spatial overlap degree, and the trajectory matching score model is as follows:

[0132]

[0133] wherein, is a weight parameter representing the object continuous stationary probability, is a weight parameter representing the smoothness score; is a weight parameter representing the spatial overlap degree. , and The sum of the weight parameters of the three is 1, which is set according to the scene, for example, when the vehicle is stationary, the image is clear and stable, and the stationary judgment is given priority, and the weight parameters are 0.6, 0.3 and 0.1 respectively, when the peak personnel is dense, the trajectory is mixed, and the smoothness score is given priority, and the weight parameters are 0.3, 0.5 and 0.2 respectively, and when the image space position information is accurate, the spatial overlap degree is given priority, and the weight parameters are 0.4, 0.2 and 0.4 respectively.

[0134] The trajectory consistency analysis method in the prior art mostly only focuses on the object position or description result of an image frame, cannot model the dynamic trajectory behavior of an object in a period of time, cannot judge whether the object is stationary or a passenger carries an object for a short stay, and has weak object motion state analysis capability, does not have the ability to analyze whether the object is stationary for a long time, especially cannot identify a non-lost object temporarily placed in the vehicle, and has high misidentification rate. In the embodiment, the trajectory consistency score is calculated by combining the stationary probability, trajectory smoothness and spatial overlap degree, the real lost object that is stationary for a long time and is not taken away, and the object carried by a person or moved in the middle are identified by using the stationary probability and trajectory smoothness information, the lost object identification rate is improved, and the matching accuracy is significantly improved by calculating the spatial overlap degree between the possible activity area of a passenger in the car compartment and the actual appearance area of a candidate object according to the ticket data and the boarding and alighting points.

[0135] In one embodiment, step S430: obtaining a lost object judgment result according to the trajectory consistency score, comprising:

[0136] Step S431: judging whether the trajectory consistency score is greater than a preset high-confidence threshold value;

[0137] Step S432: if it is judged that yes, the candidate object is judged as a high-confidence lost object, the high-confidence lost objects are summarized to generate a lost object judgment result, the candidate lost object image and description are pushed to the passenger, and the passenger is notified to come to claim.

[0138] In the embodiment, if it is judged that no, that is, the trajectory consistency score is less than the preset high-confidence threshold value, it is further judged whether it is greater than a low-confidence threshold value, at this time, it is judged as a medium-confidence lost object, the passenger is notified to come to confirm whether it is a lost object, and then, if the score is lower than the low-confidence threshold value, it is excluded from the lost object list.

[0139] In one embodiment, as shown in Figure 2 The system further comprises:

[0140] A stationary object generation module is configured to obtain target lost object information of a lost item, analyze the target lost object, and generate a stationary object set.

[0141] The lost property description generation module is configured to generate semantic image descriptions based on the stationary object set, and generate a candidate lost property description set based on a preset semantic description generation function.

[0142] The suspected target generation module is configured to perform description similarity matching based on the candidate lost property description set, and obtain a suspected target set.

[0143] The judgment result generation module is configured to perform trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model, and obtain a lost property judgment result.

[0144] In one embodiment, the stationary object generation module is further configured to: obtain target lost property information of a lost item, and generate a target segment set based on the target lost property information; perform image feature analysis based on the target segment set, and obtain a stationary object set.

[0145] In one embodiment, the stationary object generation module is further configured to: perform spatial feature vector extraction based on the target segment set, calculate spatial similarity of spatial feature vectors between frame images based on the spatial feature vectors; obtain an image matrix based on the target segment set, and generate a time stability score based on the image matrix; generate a space-time stability score of the lost object based on the spatial similarity and the time stability score; and obtain the stationary object set based on the space-time stability score.

[0146] In one embodiment, the lost property description generation module is further configured to: perform image feature extraction based on the target segment set to generate an image feature vector of the stationary object; generate a matching weight of an anchor point based on two-dimensional position coordinates of the stationary object, wherein the anchor point is pre-configured; set a semantic label of an anchor point with the largest weight as a semantic reference of the stationary object; generate a sentence description based on the image feature vector and the semantic label of the anchor point based on a semantic description generation function; and generate a candidate lost property description set based on the sentence description.

[0147] In one embodiment, the suspected target generation module is further configured to: obtain description information provided by a passenger at a lost property declaration request; perform description similarity matching on the candidate lost property description set and the description information provided at the lost property declaration request based on a preset semantic encoder, and obtain a suspected target set.

[0148] In one embodiment, the judgment result generation module is further configured to: generate an object continuous stationary probability, a stationarity score, and a spatial overlap degree based on the suspected target set; generate a trajectory consistency score based on the object continuous stationary probability, the stationarity score, and the spatial overlap degree based on a preset trajectory matching scoring model; and obtain a lost property judgment result based on the trajectory consistency score.

[0149] In one embodiment, the result generation module is further configured to: determine whether the trajectory consistency score is greater than a preset high-confidence threshold; and if the determination is yes, determine the candidate object as a high-confidence lost object, and aggregate the high-confidence lost objects to generate a lost-and-found result.

[0150] In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above-mentioned lost-and-found method based on a smart bus platform when executing the computer program.

[0151] In one embodiment, a computer readable storage medium is also provided, which stores a computer program, and the computer program implements the steps of the above-mentioned lost-and-found method based on a smart bus platform when executed by a processor.

[0152] It should be noted that the information interaction, execution process and the like between the above modules are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought by the above modules can be referred to the method embodiments part, which will not be repeated here.

[0153] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is taken as an example for illustration, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or software. In addition, the specific name of each functional unit and module is only for easy distinction, and does not limit the protection scope of the present application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0154] It should be noted that the information interaction, execution process and the like between the above modules are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought by the above modules can be referred to the method embodiments part, which will not be repeated here.

[0155] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to complete all or part of the above described functions. Each functional unit or module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit. In addition, the specific name of each functional unit or module is only for convenient distinction, and does not limit the protection scope of the present application. The specific working process of the unit or module in the system can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0156] The embodiments of the present application further provide a network device, comprising at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps in any of the above method embodiments when executing the computer program.

[0157] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program, wherein the computer program is executable by a processor to implement the steps in any of the above method embodiments.

[0158] The embodiments of the present application provide a computer program product, which, when running on a mobile terminal, enables the mobile terminal to implement the steps in any of the above method embodiments.

[0159] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can at least include any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0160] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0161] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0162] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the above-described apparatus / network device embodiments are merely schematic, for example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0163] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0164] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

[0165] An embodiment of the present application further provides a computer device, which comprises at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-mentioned embodiments.

[0166] The computer device can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above description is an example of the computer device, and does not constitute a limitation on the computer device, and can include more or fewer components than the above description, or combine certain components, or different components, for example, can also include input / output devices, network access devices, etc.

[0167] The processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0168] The memory can be an internal storage unit of the computer device in some embodiments, such as a hard disk or a memory of the computer device. The memory can also be an external storage device of the computer device in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the memory can include both an internal storage unit and an external storage device of the computer device. The memory is used to store an operating system, an application program, a boot loader, data, and other programs, such as program codes of the computer program, and the like. The memory can also be used to temporarily store data that has been output or is to be output.

[0169] The technical features of the above embodiments can be combined in any manner. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described, but it should be understood that any combination of the technical features is within the scope of the present disclosure as long as the combination does not result in a contradiction.

[0170] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the scope of the present application. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and these are within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A lost and found method based on a smart bus platform, characterized in that, The method comprises: obtaining target lost article information of a lost article, analyzing the target lost article, and generating a stationary object set; generating semantic image description based on the stationary object set, and generating a candidate lost article description set based on a preset semantic description generation function; performing description similarity matching based on the candidate lost article description set to obtain a suspected target set; performing trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model to obtain a lost article judgment result, comprising: Generating an object stationary probability from the set of suspect targets , stationarity score , and spatial overlap ; Object stationary probability representing an object stationary probability in an object loss time period; Smoothness score represents a smoothness score of the object trajectory over the object loss time period; spatial overlap representing a spatial overlap of the passenger trajectory area with the object appearance location; Generating a stationary probability of the object comprising: ； wherein, is the probability of the object staying still, is the spatial trajectory sequence of the object, including time stamp and two-dimensional position coordinates; is the m-th frame image in the object loss time period; is the object loss time period described by the passenger; is the mode of the object position in the object loss time period, i.e. the most frequently occurring position; is the fluctuation of the object position in the object loss time period, if the distance between the spatial position of the object in each frame image in the time period and the mode of the object position is less than the set threshold, i.e. the position fluctuation tolerance range, then the value is 1, otherwise the value is 0; is the total number of frame images in the object loss time period; represents that the center coordinate of the object in the m-th frame image in the object loss time period belongs to ; and the m-th frame image is in the time period. Generating stationarity scores As follows: ； wherein the smoothness control factor is used to normalize the square of the calculated inter-object Euclidean distance to a dimensionless number, which is set according to the tolerance level of trajectory fluctuation, , is the tolerance of trajectory fluctuation, in units of pixels, the higher the tolerance level of trajectory fluctuation, the higher the smoothness score under the same trajectory fluctuation; : the two-dimensional position coordinates of the object in the previous frame image of the object; : the square of the Euclidean distance of the object between two frame images, reflecting the movement of the object; represents the square value of the average inter-frame movement distance of the object over the time period, if the object is stationary, the value is equal to or close to 0, if the object is constantly moving, the value changes significantly; the square of the Euclidean distance is a geometric standard of two-dimensional translation, which is direction-independent; The spatial overlap degree The proportion of the passenger seating area and the object appearing area is calculated, and the specific calculation formula is as follows: ； wherein is the number of regions in which the object appears and the passenger activity region overlaps, obtained from the number of regions in which the object is captured in the passenger activity region; is the number of passenger activity regions; generating a trajectory consistency score based on the object continuous stationary probability, the smoothness score, and the spatial overlap degree based on the preset trajectory matching scoring model; The trajectory matching scoring model is as follows: ； wherein, is a weight parameter representing the probability of the object remaining stationary, is a weight parameter representing the stationarity score; is a weight parameter representing the spatial overlap; , and the sum of the weight parameters is 1. obtaining a lost article judgment result based on the trajectory consistency score. 2.The lost and found method based on the smart bus platform of claim 1, wherein, Obtaining target lost article information of a lost article, analyzing the target lost article, and generating a stationary object set, comprising: obtaining target lost article information of a lost article, and generating a target segment set based on the target lost article information; performing image feature analysis based on the target segment set to obtain a stationary object set. 3.The lost and found method based on the smart bus platform of claim 2, wherein, Performing image feature analysis based on the target segment set to obtain a stationary object set, comprising: extracting a spatial feature vector based on the target segment set, and calculating the spatial similarity of the spatial feature vectors between frame images based on the spatial feature vector; obtaining an image matrix based on the target segment set, and generating a time stability score based on the image matrix; generating a spatiotemporal stability score of the lost object based on the spatial similarity and the time stability score; obtaining a stationary object set based on the spatiotemporal stability score. 4.The lost and found method based on the smart bus platform of claim 2, wherein, Generating semantic image description based on the stationary object set, and generating a candidate lost article description set based on a preset semantic description generation function, comprising: extracting an image feature vector of a stationary object based on the target segment set; generating a matching weight of an anchor point based on the two-dimensional position coordinates of the stationary object, wherein the anchor point is pre-configured; setting the semantic label of the anchor point with the largest weight as the semantic reference of the stationary object; generating a sentence description based on the image feature vector and the semantic label of the anchor point based on the semantic description generation function, generating a candidate lost article description set based on the sentence description. 5.The lost and found method based on the smart bus platform of claim 1, wherein, Performing description similarity matching based on the candidate lost article description set to obtain a suspected target set, comprising: obtaining description information provided by a passenger at the time of lost article declaration request; performing description similarity matching based on a preset semantic encoder to obtain a suspected target set. 6.The lost and found method based on the smart bus platform of claim 1, wherein, Obtaining a lost article judgment result based on the trajectory consistency score, comprising: determining whether the trajectory consistency score is greater than a preset high-confidence threshold; if the determination is yes, determining that the candidate object is a high-confidence lost article, and generating a lost article judgment result by aggregating each high-confidence lost article.

7. A lost and found system based on a smart bus platform, used to realize the lost and found method based on the smart bus platform in claim 1, characterized in that, The system comprises: a stationary object generation module for obtaining target lost article information of a lost article, analyzing the target lost article, and generating a stationary object set; a lost article description generation module for generating semantic image description based on the stationary object set, and generating a candidate lost article description set based on a preset semantic description generation function; A suspected target generation module is configured to perform description similarity matching based on the candidate lost property description set, and obtain a suspected target set; A judgment result generation module is configured to perform trajectory consistency analysis on the suspected target set based on a preset trajectory matching scoring model, and obtain a lost property judgment result.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 6.

Images