CN114218992A

CN114218992A - Abnormal object detection method and related device

Info

Publication number: CN114218992A
Application number: CN202111643714.0A
Authority: CN
Inventors: 陈方云; 夏凤君; 汪昊; 周斌
Original assignee: Chongqing Unisinsight Technology Co Ltd
Current assignee: Chongqing Unisinsight Technology Co Ltd
Priority date: 2021-12-29
Filing date: 2021-12-29
Publication date: 2022-03-22
Anticipated expiration: 2041-12-29
Also published as: CN114218992B

Abstract

The application relates to the technical field of computers, and discloses a method and a related device for detecting an abnormal object, which are used for solving the problems of low flexibility and poor accuracy of abnormal personnel early warning generated by related technologies. According to the method and the device, the prior probability table is established in advance, the characteristics of the abnormal object are extracted from the collected data based on the prior probability table, the prior probability corresponding to the extracted characteristics is searched from the prior probability table, and then the extracted characteristics and the prior probability are processed by adopting a Bayesian probability model, so that which objects in the collected data are the abnormal objects for alarming can be predicted, the real-time effectiveness and accuracy of early warning are ensured, and the difficulty of manual checking is reduced. And the abnormal object is detected by acquiring the historical moving track of the candidate object and predicting the confidence coefficient of the next moving track through the neural network model, so that the abnormal object is verified with the abnormal object predicted by adopting the Bayesian probability model, and the early warning accuracy of the abnormal personnel is further ensured.

Description

Abnormal object detection method and related device

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method and a related apparatus for detecting an abnormal object.

Background

With the rapid development of social infrastructure construction, in order to maintain social security, the analysis and early warning requirements based on abnormal personnel are increasing day by day.

At present, early warning of a plurality of abnormal behaviors is basically generated by a front-end camera in order to ensure real-time performance, but due to the limitations of calculation power of the camera and early warning rules, the generated abnormal warning has a plurality of false alarms and needs manual real-time intervention and verification. However, in practical application, it is difficult to exhaust the early warning rules of various changing conditions, which results in poor flexibility and low accuracy of the method.

The other early warning mode is to adopt an integral model, coefficients of various regular characteristics of the integral model are mainly set by empirical values, certain flexibility is lacked, and the generated early warning accuracy of abnormal personnel is poor.

Disclosure of Invention

The application provides a detection method and a related device for an abnormal object, which are used for solving the problems of low flexibility and poor accuracy of abnormal personnel early warning generated by related technologies.

In a first aspect, an embodiment of the present application provides a method for detecting an abnormal object, including:

acquiring alarm information of an abnormal object, wherein the alarm information comprises a sensor position sending the alarm information;

acquiring data acquired by a sensor to be processed around the position of the sensor by taking the position of the sensor as a reference;

extracting characteristic information of a candidate object from data collected by a sensor to be processed, wherein the characteristic information comprises multi-dimensional characteristics;

acquiring prior probability corresponding to the feature of each dimension in the multi-dimensional features from a preset prior probability table; wherein the prior probabilities include a prior probability that a candidate is the anomalous object and a prior probability that the candidate is not the anomalous object;

processing the multi-dimensional features and the prior probability corresponding to the feature of each dimension by adopting a Bayesian probability model to obtain a risk coefficient of the candidate object; wherein the risk factor is indicative of a probability that the candidate object is the abnormal object;

and if the risk coefficient of the candidate object is greater than or equal to a preset risk coefficient threshold value, determining that the candidate object is the abnormal object.

In a possible implementation manner, if it is determined that the risk factor of the candidate object is greater than or equal to a preset risk factor threshold and before it is determined that the candidate object is the abnormal object, the method further includes:

extracting a moving track of the candidate object from data collected by a plurality of sensors;

inputting the moving track into a neural network model to obtain the confidence coefficient of the next moving track of the candidate object output by the neural network model, wherein the confidence coefficient is used for representing the probability that the next moving track is the moving track of a normal object;

determining that the confidence of the candidate object is less than a preset confidence threshold.

In one possible implementation, the multi-dimensional features include:

the method comprises the following steps of performing image matching on a candidate object, wherein the image matching comprises space-time characteristics of the candidate object, appearance characteristics of the candidate object, behavior characteristics of the candidate object, confidence degree comparison characteristics and association target characteristics, the association target characteristics are used for representing that a motor vehicle is extracted from a specified range of the position of the candidate object within a specified time length, the confidence degree comparison characteristics comprise the similarity confidence degree of the candidate object and an abnormal object, and/or the image quality of the candidate object is lower than an image quality score.

In a possible implementation manner, before the obtaining, from a preset prior probability table, a prior probability corresponding to a feature of each dimension in the multi-dimensional features, the method further includes:

determining that the features of the candidate object do not match the features of the objects in the whitelist.

In a possible implementation manner, the extracting the movement trajectory of the candidate object from the data acquired by the plurality of sensors specifically includes:

acquiring data of the candidate objects acquired by a plurality of sensors;

and constructing a movement track of the candidate object based on the time information in the data of the candidate object and the position information of each sensor.

In one possible embodiment, training the neural network model comprises:

obtaining a training sample, wherein the training sample comprises a historical moving track of a sample object, a next moving track of the historical moving track and a target confidence coefficient;

inputting the historical movement track of the sample object into the neural network model to obtain the next movement track and the prediction confidence of the sample object output by the neural network model;

comparing the next moving track output by the neural network model with the next moving track included in the training sample to obtain position loss, and determining confidence loss based on the prediction confidence and the target confidence;

training the neural network model using the position loss and the confidence loss.

In one possible embodiment, the method further comprises:

and if the candidate object is determined to be the abnormal object, determining that the sensor to be processed is used as a reference, and returning to execute the step of acquiring data collected by the sensor to be processed around the position of the sensor by using the position of the sensor as a reference.

In a second aspect, an embodiment of the present application provides an apparatus for detecting an abnormal object, where the apparatus includes:

the alarm position acquisition module is used for acquiring alarm information of an abnormal object, wherein the alarm information comprises a sensor position for sending the alarm information;

the data acquisition module is used for acquiring data acquired by the sensor to be processed around the position of the sensor by taking the position of the sensor as a reference;

the multi-dimensional feature extraction module is used for extracting feature information of candidate objects from data collected by a sensor to be processed, wherein the feature information comprises multi-dimensional features;

the prior probability obtaining module is used for obtaining the prior probability corresponding to the feature of each dimension in the multi-dimensional features from a preset prior probability table; wherein the prior probabilities include a prior probability that a candidate is the anomalous object and a prior probability that the candidate is not the anomalous object;

a risk coefficient determining module, configured to process the multidimensional features and the prior probability corresponding to the feature of each dimension by using a bayesian probability model to obtain a risk coefficient of the candidate object; wherein the risk factor is indicative of a probability that the candidate object is the abnormal object;

and the abnormality identification module is used for determining the candidate object as the abnormal object if the risk coefficient of the candidate object is greater than or equal to a preset risk coefficient threshold value.

In a possible implementation manner, if it is determined that the risk factor of the candidate object is greater than or equal to a preset risk factor threshold and before it is determined that the candidate object is the abnormal object, the apparatus further includes:

the moving track extraction module is used for extracting the moving tracks of the candidate objects from data collected by a plurality of sensors;

a confidence coefficient obtaining module, configured to input the movement trajectory into a neural network model, to obtain a confidence coefficient of a next movement trajectory of the candidate object output by the neural network model, where the confidence coefficient is used to represent a probability that the next movement trajectory is a movement trajectory of a normal object;

and the anomaly identification module is specifically used for determining that the confidence coefficient of the candidate object is smaller than a preset confidence coefficient threshold value.

In a possible implementation, the multidimensional feature extracted by the multidimensional feature extraction module includes:

In a possible implementation manner, before the prior probability obtaining module is configured to obtain the prior probability corresponding to the feature of each dimension in the multi-dimensional feature from a preset prior probability table, the apparatus further includes:

and the characteristic matching module is used for determining that the characteristics of the candidate object are not matched with the characteristics of the objects in the white list.

In a possible implementation, the extracting of the movement trajectory of the candidate object from the data acquired by the plurality of sensors is performed, and the movement trajectory extracting module is specifically configured to:

acquiring data of the candidate objects acquired by a plurality of sensors;

In one possible embodiment, the apparatus comprises:

the training sample acquisition module is used for acquiring a training sample, wherein the training sample comprises a historical moving track of a sample object, a next moving track of the historical moving track and a target confidence coefficient;

the confidence coefficient acquisition module is further used for inputting the historical movement track of the sample object into the neural network model to obtain a next movement track and a prediction confidence coefficient of the sample object output by the neural network model;

a loss determination module, configured to compare a next movement trajectory output by the neural network model with a next movement trajectory included in the training sample to obtain a position loss, and determine a confidence loss based on the prediction confidence and the target confidence;

a training module to train the neural network model using the location loss and the confidence loss.

In one possible embodiment, the apparatus comprises:

and the iteration module is used for determining that the sensor to be processed is used as a reference if the candidate object is determined to be the abnormal object, and returning to execute the step of acquiring data acquired by the sensor to be processed around the sensor position by using the sensor position as a reference.

In a third aspect, an embodiment of the present application provides an electronic device, including:

a processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the method of detecting any abnormal object as provided in the first aspect above.

In a fourth aspect, the present application further provides a computer-readable storage medium, where instructions, when executed by a processor of an electronic device, enable the electronic device to perform any one of the above-mentioned abnormal object detection methods provided in the first aspect.

In a fifth aspect, the present application provides a computer program product, which includes a computer program, and the computer program is executed by a processor to implement any one of the abnormal object detection methods provided in the first aspect.

The technical scheme provided by the embodiment of the application at least has the following beneficial effects:

the embodiment of the application summarizes various dimensional characteristics of the abnormal object in the table by pre-establishing the prior probability table. And after the abnormal object invades, a sensor is triggered to alarm, and then data collected by each sensor is searched in a prevention and control circle taking the sensor as a reference. Then, the characteristics of the abnormal object are extracted from the collected data based on the prior probability table, the prior probability corresponding to the extracted characteristics is searched from the prior probability table, and then the extracted characteristics and the prior probability are processed by adopting a Bayesian probability model, so that which objects in the collected data are the alarming abnormal objects can be predicted. Therefore, the multi-dimensional characteristics of the candidate object are automatically extracted, and the prior probability corresponding to the characteristics is input into the Bayesian probability model by utilizing the algorithm, so that the abnormal object can be detected, the abnormal object is early warned, the real-time effectiveness and accuracy of early warning are further ensured, and the difficulty of manual checking is reduced.

In addition, the method and the device further acquire the historical moving track and the next moving track of the sample object, train the neural network model to obtain the confidence coefficient of the next moving track output by the neural network model and the next moving track in the training sample, predict the confidence coefficient of the next moving track of the candidate object through the trained neural network model according to the historical moving track of the candidate object, and mutually verify the confidence coefficient of the next moving track of the candidate object, so that the early warning accuracy of abnormal personnel is further ensured.

Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic view of an application scenario of a method for detecting an abnormal object according to an embodiment of the present application;

fig. 2 is a schematic flowchart of a method for detecting an abnormal object according to an embodiment of the present application;

fig. 3 is a schematic flowchart of another abnormal object detection method according to an embodiment of the present application;

fig. 4 is a schematic diagram of a movement track of a candidate object according to an embodiment of the present disclosure;

fig. 5 is a schematic flowchart of a training method of a neural network model according to an embodiment of the present disclosure;

fig. 6 is a schematic diagram illustrating a motion trajectory prediction of a candidate object according to an embodiment of the present disclosure;

fig. 7 is a schematic structural diagram of an apparatus for detecting an abnormal object according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. The embodiments described are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Hereinafter, some terms in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.

(1) The terms "first" and "second" in the embodiments of the present application are used for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", may explicitly or implicitly include one or more features, and in the description of embodiments of the application, a "plurality" means two or more unless stated otherwise.

(2) "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. Also, in the description of the embodiments of the present application, unless otherwise specified, the character "/" generally indicates that the former and latter associated objects are in an "or" relationship, for example, a/B may indicate a or B.

(3) A server serving the terminal, the contents of the service such as providing resources to the terminal, storing terminal data; the server is corresponding to the application program installed on the terminal and is matched with the application program on the terminal to run.

(4) The terminal may refer to an APP (Application) of a software class or a client. The system is provided with a visual display interface and can interact with a user; is corresponding to the server, and provides local service for the client. For software applications, except some applications that are only run locally, the software applications are generally installed on a common client terminal and need to be run in cooperation with a server terminal. After the internet has developed, more common applications include e-mail clients for e-mail receiving and sending, and instant messaging clients. For such applications, a corresponding server and a corresponding service program are required in the network to provide corresponding services, such as database services, configuration parameter services, and the like, so that a specific communication connection needs to be established between the client terminal and the server terminal to ensure the normal operation of the application program.

Any number of elements in the drawings are by way of example and not by way of limitation, and any nomenclature is used solely for differentiation and not by way of limitation.

In the related art, the early warning of the abnormal person is basically generated by a front-end camera through an early warning rule or an integral model. Aiming at the early warning rules, different early warning rules need to be formulated according to different scenes so as to realize early warning of abnormal personnel. For example: the information of abnormal personnel needs to be imported for list distribution and control, or the detection is carried out through a human behavior detection algorithm, and the like. The human behavior detection algorithm is as follows: regional loitering and regional intrusion detection. And then the front-end equipment generates early warning according to the early warning rule.

The integral model calculates the comprehensive score by configuring weights of different characteristics, and the mode cannot adapt to the condition that the characteristics need to be bundled and scored and the weights need to be automatically updated in incremental change, so that the generated early warning accuracy of abnormal personnel is poor and manual real-time intervention and verification are needed.

The early warning rules are difficult to exhaust various change conditions, and the integral model needs to set characteristic coefficients by means of empirical values, so that the early warning by using the early warning rules and the early warning by using the integral model lack certain flexibility and are low in accuracy. The method aims to improve the real-time performance and effectiveness of the early warning of abnormal personnel and quickly and effectively provide effective early warning for the abnormal personnel with abnormal behaviors, so that real-time intervention can be quickly and effectively carried out, and the method is a technical problem which needs to be solved urgently at present.

In view of this, the present application provides a method and a related device for detecting an abnormal object, which are used to solve the problems of poor flexibility and accuracy of abnormal personnel early warning and the need of manual real-time intervention and verification in related technologies.

The invention conception of the invention is as follows: a prior probability table is established in advance. The table summarizes various dimensional characteristics of the anomaly object. And after the abnormal object invades, a sensor is triggered to alarm, and then data collected by each sensor is searched in a prevention and control circle taking the sensor as a reference. Then, the characteristics of the abnormal object are extracted from the collected data based on the prior probability table, the prior probability corresponding to the extracted characteristics is searched from the prior probability table, and then the extracted characteristics and the prior probability are processed by adopting a Bayesian probability model, so that which objects in the collected data are the alarming abnormal objects can be predicted. Therefore, the abnormal objects of early warning can be detected by extracting the multi-dimensional characteristics of the candidate objects and adopting the Bayesian probability model, so that the real-time effectiveness and accuracy of the early warning are ensured, and the difficulty of manual checking is reduced.

In addition, the confidence coefficient of the next moving track and the risk coefficient of the candidate object are verified mutually by acquiring the historical moving track of the candidate object and predicting the confidence coefficient of the next moving track through the neural network model, so that the accuracy of early warning of abnormal personnel is further ensured.

After the inventive concept of the embodiment of the present application is introduced, some simple descriptions are made below on application scenarios to which the technical solution of the embodiment of the present application can be applied, and it should be noted that the application scenarios described below are only used for describing the embodiment of the present application and are not limited. In specific implementation, the technical scheme provided by the embodiment of the application can be flexibly applied according to actual needs.

Fig. 1 is a schematic view of an application scenario of the method for detecting an abnormal object according to the embodiment of the present application. The application scenario includes: sensors 101, sensors 102, sensors 103, sensors 104, network 105, server 106, memory 107. The sensor provided by the embodiment of the application can be an image sensor, a radar sensor, an electronic fence sensor and other devices capable of acquiring data of abnormal objects, and the sensor is suitable for the embodiment of the application. The number and type of the sensors are not limited in the embodiments of the present application.

The server 106 first acquires alarm information of an abnormal object sent by the sensor 101, acquires the position of the sensor 101, acquires data collected by the sensors 102, 103, and 104 around the position of the sensor 101 with reference to the position of the sensor 101, detects the probability that the candidate object is an abnormal object in the sensor 101 based on the data of the candidate object collected by the sensors 102, 103, and 104, and sends an early warning if the candidate object is detected to be an abnormal object in the sensor 101.

The sensors 101, 102, 103, 104 and the server 106 are connected through a wireless or wired network 105, and the server 106 may be a server, a server cluster formed by a plurality of servers, or a cloud computing center. The server 106 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a network service, cloud communication, middleware service, a domain name service, a security service, a CDN, a big data and artificial intelligence platform, and the like.

Of course, the method provided in the embodiment of the present application is not limited to be used in the application scenario shown in fig. 1, and may also be used in other possible application scenarios, and the embodiment of the present application is not limited. The functions that can be implemented by each device in the application scenario shown in fig. 1 will be described in the following method embodiments, and will not be described in detail herein.

To further illustrate the technical solutions provided by the embodiments of the present application, the following detailed description is made with reference to the accompanying drawings and the detailed description. Although the embodiments of the present application provide the method operation steps as shown in the following embodiments or figures, more or less operation steps may be included in the method based on the conventional or non-inventive labor. In steps where no necessary causal relationship exists logically, the order of execution of the steps is not limited to that provided by the embodiments of the present application.

Referring to fig. 2, a schematic flowchart of a method for detecting an abnormal object according to an embodiment of the present application is shown. As shown in fig. 2, the method comprises the steps of:

in step 201, alarm information of an abnormal object is acquired, wherein the alarm information comprises a sensor position sending the alarm information.

In a possible implementation manner, if an abnormal object is captured by a sensor in the embodiment of the present application, an alarm prompt is performed, and alarm information corresponding to the abnormal object is generated. For example, a vibrating optical fiber is arranged in the border wall, if a person jumps over the border wall, the vibrating optical fiber in the border wall triggers vibration to send an alarm prompt, and then the equipment associated with the vibrating optical fiber captures the morphological characteristics of the person who rolls over the wall, so that the corresponding data characteristics are obtained.

The alarm information may include the appearance characteristics, behavior characteristics, etc. of the abnormal object, and also include the position of the sensor giving the alarm prompt. Where the sensor may be a radar, electronic fence, surveillance camera, or other sensing device that can sense an abnormal person.

In step 202, data collected by the sensors to be processed is acquired around the sensor position with reference to the sensor position.

In a possible implementation manner, after the alarm information of the abnormal object is obtained, the abnormal object needs to be searched and tracked, and in order to determine the search range, in this embodiment of the application, a prevention and control circle of the dynamic sensitive area may be generated based on the position of the sensor which sends the alarm information obtained in step 201, and then data acquired by a sensor to be processed in the prevention and control circle may be obtained.

For example, after receiving the alarm prompt, a dynamic prevention and control circle composed of various sensors is formed by taking the sensor sending the alarm prompt as the center and taking the distance in which the abnormal object with the preset duration can move as the radius, and then the abnormal object is continuously searched and tracked in all the sensors to be processed in the dynamic prevention and control circle. The preset duration is the duration from the time when the alarm prompt is received to the time when the data acquired by the sensor is currently acquired, the preset duration can be dynamically updated according to the time when the data acquired by the sensor is acquired, and the longer the time when the data acquired by the sensor is far away from the time when the alarm prompt is received, the longer the preset duration is.

The dynamic prevention and control ring in the embodiment of the application can be understood as the prevention and control ring of different sensors according to the alarm prompt sequence. For example, if the sensor a gives an alarm first, an abnormal object is searched in the control circle of the sensor a, and then, if an abnormal object is searched in the sensor B in the control circle of the sensor a, the abnormal object is continuously searched in the control circle of the sensor B. Therefore, the prevention and control ring moves from the prevention and control ring of the sensor A to the prevention and control ring of the sensor B, and the dynamic change of the prevention and control ring is realized.

For example, after a person is found to jump a border wall, a nearby camera and other sensing devices generate an abnormal alarm, then a preset time length is assumed to be set to be 15 minutes by taking an alarm point as a center, or the total time length of 15 minutes plus an expected search time length is taken as the preset time length, the preset time length is multiplied by the normal speed of the person to obtain a moving distance, then a prevention and control ring composed of front-end monitoring devices is formed by taking the moving distance as a radius, and then the abnormal object is continuously searched and tracked in all the monitoring devices in the dynamic prevention and control ring by taking the prevention and control ring as a reference.

Taking a control circle as an example, in step 203, feature information of the candidate object is extracted from the data collected by the sensor to be processed, where the feature information includes multi-dimensional features.

In one possible implementation, the multi-dimensional features extracted in the embodiment of the present application include a spatiotemporal feature of the candidate object, an appearance feature of the candidate object, a behavior feature of the candidate object, a confidence level comparison feature, and an association target feature, the association target feature is used for characterizing that the motor vehicle is extracted from a specified range of the position of the candidate object within a specified duration, the confidence level comparison feature includes a confidence level that the candidate object and the abnormal object are similar, and/or the image quality of the candidate object is lower than an image quality score. For ease of understanding, these features are described below in conjunction with Table 1:

TABLE 1

1. Spatiotemporal features of candidates

As the name implies, the spatiotemporal features of a candidate are feature information of the candidate extracted from the sensor to be processed in time and space.

In a possible implementation mode, the abnormal object may have the behavior of hiding the position information, that is, after the sensors in the control circle reach a certain coverage degree, the abnormal object is likely to appear in the sensors with small occurrence probability of the normal object, so that the space-time characteristics of the candidate object need to be extracted. As shown in table 1, the spatiotemporal features may include features that a candidate object extracted from the sensor to be processed may arrive at the sensor to be processed from the last position in a spatiotemporal space, and the position of the sensor to be processed is a position where the probability of occurrence of a normal object is small. The last position can be understood as the last moving track point of the abnormal object, and can also be understood as the last sensor position for detecting the abnormal object.

In a possible implementation manner, in order to determine the spatio-temporal characteristics, that is, in order to determine that a candidate object extracted from a sensor to be processed can reach the sensor to be processed from the last position in the time space and that the position of the sensor to be processed is a position where the probability of occurrence of a normal object is very small, in the embodiment of the present application, it is further required to obtain the position environment of the candidate object in the sensor to be processed, including mountainous regions, farmlands, paths, roads, and the like, as well as the position information, including geographical position information, information associated with front and back sensors to be processed, and the running/walking time of the candidate object passing through two adjacent sensors to be processed according to the time sequence. The information related to the front and rear to-be-processed sensors includes position information related to two adjacent to-be-processed sensors through which the candidate object passes, position environment information acquired by the two adjacent to-be-processed sensors, and the like.

The running/walking time of the acquired candidate object passing through the two sensors to be processed and the path distance between the two sensors are used for calculating the speed of the candidate object required for passing through the two sensors, so that whether the candidate object extracted from the sensors to be processed can reach the sensors to be processed from the last position in time space or not can be determined. For example, the calculated speed is greater than 10 m/sec, which is impossible to reach for the average walking speed of the person, it may be determined that the candidate object extracted from the sensor to be processed may not reach the sensor to be processed from the last position in the time space. And determining whether the geographical position is a position where the probability of occurrence of the normal object is small by acquiring the position information of the candidate object. For example, the position information of the candidate object is in a forest, it can be determined that the position of the candidate object is a position where the probability of occurrence of the normal object is small.

2. Confidence level comparison characteristic of candidate object

In the embodiment of the present application, the confidence characteristic may be that the confidence of similarity between the candidate object and the abnormal object is higher, and higher confidence may be higher than a certain confidence threshold. The confidence degree comparison features comprise similar confidence degrees of the candidate object and the abnormal object and/or image quality of the candidate object is lower than an image quality score, for example, a human face of the candidate object extracted from the sensor to be processed has a higher confidence degree with a human face of the abnormal object, a human body of the candidate object has a higher confidence degree with a human body of the abnormal object, and a picture of the candidate object captured by the sensor to be processed has a lower picture quality score. The confidence degree of the similarity between the candidate object and the abnormal object is influenced by the image quality score of the image of the sensor to be processed capturing the candidate object, the lower the image quality score of the image capturing the candidate object is, the more fuzzy the candidate object is, the lower the definition is, the more difficult the confidence degree of the similarity between the candidate object and the abnormal object is to be judged, and at the moment, the candidate object may be an abnormal person.

The method comprises the steps that when a sensor sends alarm information of an abnormal object, some obvious features of the abnormal object, such as human faces, human body information and picture quality scores of a snapshot picture, are collected, then the information is uploaded to a server to be stored, when the sensor to be processed extracts information such as the human faces, the human bodies and the picture quality scores of the snapshot picture of candidate objects, the information is also uploaded to the server, a confidence coefficient threshold value can be set in the server in advance, the information of the abnormal object and the information of the candidate objects are compared, corresponding confidence coefficients can be obtained, and if the confidence coefficients are larger than the confidence coefficient threshold value, corresponding prior probabilities corresponding to the confidence coefficient comparison features are inquired from a table 1. The method for obtaining the confidence coefficient may be directly calculating the pixel value, may also be calculating according to a face contrast model, and may also be calculating using a neural network model, which is not limited in the embodiment of the present application.

3. Appearance features of candidate objects

In one possible embodiment, the appearance features include hat wearing, sunglasses wearing, mask wearing, slippers wearing, cat waist shape, fast moving standby processing sensors extracted features that can obviously describe the appearance of the candidate object.

The sensor sends out alarm information of the abnormal object and collects some obvious appearance characteristics of the abnormal object, such as information of wearing black clothes, wearing sunglasses and the like, then the information is uploaded to the server to be stored, the information is also uploaded to the server after the sensor to be processed extracts picture information of the candidate object, and if a processor in the server detects the appearance information of the candidate object, which is the same as that of the abnormal object, the processor in the server can inquire the corresponding prior probability corresponding to the appearance characteristics from the table 1.

4. Behavioral characteristics of candidate objects

In one possible embodiment, the behavior feature is used to characterize the behavior consistency of the candidate correspondences of the sensors at the strip with the behavior of the anomalous object of the last sensor. In implementation, the behavior consistency can be represented by a smaller included angle between the position of the candidate object extracted from the sensor to be processed and the position of the sensor sending the alarm information.

In order to determine the behavior characteristics, that is, to determine that a smaller included angle exists between the position of the candidate object extracted from the sensor to be processed and the position of the sensor sending the alarm information, in the embodiment of the present application, an included angle threshold may be set in advance, and at the same time, the position information of the candidate object in the sensor to be processed and the position information of the sensor sending the alarm information need to be acquired, and then, based on the longitude and latitude information of the candidate object and the longitude and latitude information of the sensor sending the alarm information, an included angle between the position of the candidate object extracted from the sensor to be processed and the position of the sensor sending the alarm information may be calculated, and if the calculated included angle is smaller than the included angle threshold, the prior probability corresponding to the behavior characteristics may be queried from table 1. The smaller the included angle between the position of the candidate object extracted from the sensor to be processed and the position of the sensor giving out the alarm information is, the more consistent the advancing directions of the candidate object and the abnormal object are.

5. Associated target features of candidate objects

In one possible embodiment, the associated target feature is used to characterize a motor vehicle that occurs for a specified duration in a specified range of candidate object positions extracted from the sensor to be processed.

In order to determine the associated target features, that is, to determine the motor vehicle which appears in the specified time length in the specified range of the candidate object position extracted from the sensors to be processed, in the embodiment of the application, a specified time length and a specified range may be set, then multiple frames of images with candidate objects appearing in the specified time length are collected from the multiple sensors to be processed, and then the motor vehicle which appears in the specified range of the candidate objects is detected from the multiple frames of images. If a motor vehicle is detected, the prior probabilities corresponding to the associated target features may be queried from table 1. For example, motor vehicles that appear within 20 meters of the candidate within 15 minutes of the candidate may be detected from a plurality of sensors to be processed. The probability that the candidate object is an abnormal person is higher than the probability that the candidate object is a normal person if the motor vehicle is inquired by determining the related target characteristics, namely inquiring the motor vehicle with the occurrence of specified duration in the specified range of the candidate object position extracted from the sensor to be processed.

The embodiments of the present application include, but are not limited to, the features in table 1, and may also be used in combination to increase and decrease the multidimensional features according to actual situations and empirical values.

In one possible embodiment, in order to reduce the range of the candidate object, the candidate object whose feature does not match the feature of the object in the white list may be used as the final candidate object according to the feature comparison of the object in the white list. The features used in matching may be features in the above-mentioned multidimensional features, or may be a white list determined according to other methods. For example, objects frequently appearing in the sensor to be processed may be determined as a white list, and objects appearing for the first time or blurred shadows may be determined as candidate objects considering that the objects do not match with the features of the objects in the white list. If the features used in matching are not the features in the multidimensional features, the step of determining the candidate object by comparing the features with the objects in the white list may be performed before step 203, that is, after the candidate object is determined by filtering through the white list, the multidimensional features of the candidate object are extracted. Therefore, the workload of extracting the multi-dimensional features of the candidate objects can be reduced, and the efficiency is improved.

In step 204, obtaining a prior probability corresponding to each dimension of the multi-dimensional features from a preset prior probability table; wherein the prior probabilities include a prior probability that the candidate is an anomalous object and a prior probability that the candidate is not an anomalous object.

In a possible implementation manner, if the prior probability corresponding to each dimension feature in the prior probability table is already stable, directly obtaining the prior probability corresponding to each dimension feature in the prior probability table; if the implementation process lacks positive sample data of abnormal object model training, as shown in table 1, prior probabilities corresponding to the features of each dimension in the multi-dimensional features may be set in a prior probability table according to empirical values, and then the prior probabilities that the candidate objects corresponding to each dimension feature in the prior probability table are abnormal objects are updated by presetting time intervals, based on the number of the abnormal objects detected in the preset time corresponding to each dimension feature, the number of the candidate objects and the number of all the objects in the anti-vacancy ring; and updating the prior probability that the candidate object corresponding to the feature of each dimension in the prior probability table is not the abnormal object based on the number of normal persons in the preset time length corresponding to the feature of each dimension, the number of candidate objects and the number of all objects in the anti-vacancy ring. The prior probability corresponding to each dimension feature in the prior probability table is gradually stabilized after being continuously updated and verified.

In step 205, a bayesian probability model is used to process the multidimensional characteristics and the prior probability corresponding to the characteristics of each dimension, so as to obtain a risk coefficient of the candidate object; wherein the risk factor is used to indicate the probability that the candidate object is an abnormal object.

For example, for convenience of representation, a may be set to indicate that the candidate object and the abnormal object are the same person, B may be set to indicate that the candidate object and the abnormal object are not the same person, and P may be set to_ARepresents the probability that the candidate object is the same person as the abnormal object, and P_BThe probability that the candidate and the abnormal object are not the same person is shown, and if some candidate detects the occurrence of some of the features in the above table 1, it is respectively shown as f₁、f₂、…、f_i…,f_NThen F ═ F can be used₁，f₂、…,f_i…,f_N) Representing a feature set of the candidate object, and then using a naive bayes model, calculating the probability that the candidate object is an abnormal object based on the prior probability of each feature in the feature set using formula (1) as follows:

wherein P (F | A) represents the prior probability that the feature set F is an abnormal object, and is used

Calculated to obtain, wherein P (f)_i| a) is the prior probability that feature i is an anomalous object. P (F | B) represents the prior probability that the feature set F is a normal object, and the method is used

Calculated to obtain, wherein P (f)_i| B) is the prior probability that feature i is a normal object. For example, if a candidate object wears hat and sunglasses near a border wall and identifies a cat waist shape, and the feature set F is hat wearing, sunglasses wearing, and cat waist shape, table 1 can obtain the prior probability P (fa) of an abnormal object being the feature set a6 a7 a10, and the prior probability P (fb) of a normal object being the feature set B6B 7B 10, so that the trained naive bayesian model can obtain the probability P (af) of the candidate object being the abnormal object.

Wherein P is_AAnd P_BThe parameters are finally determined by inputting the prior probability of each feature into a naive Bayes model and continuously training.

Therefore, the probability that the candidate object is an abnormal object can be obtained through a naive Bayes model, namely, the risk coefficient of the candidate object can be determined.

In step 206, if the risk factor of the candidate object is greater than or equal to the preset risk factor threshold, the candidate object is determined to be an abnormal object. For example, a preset risk coefficient threshold value may be set to be 0.6, and if the risk coefficient of the candidate object is 0.6-1, it indicates that the candidate object is an abnormal object.

In a possible implementation manner, in order to more accurately detect whether a candidate object is an early warning object, in the embodiment of the present application, after determining that a risk coefficient of the candidate object is greater than or equal to a preset risk coefficient threshold and before determining that the candidate object is an abnormal object, the confidence of the candidate object may also be determined, so that the risk coefficient of the candidate object and the confidence of the candidate object are used to perform mutual verification, thereby more determining that the candidate object is an early warning object, and determining the confidence of the candidate object may be specifically performed as the steps shown in fig. 3:

in step 301, a movement trajectory of a candidate object is extracted from data collected by a plurality of sensors.

In a possible implementation manner, data of candidate objects acquired by a plurality of sensors may be acquired in the embodiment of the present application; based on the time information in the data of the candidate object and the position information of each sensor, a movement locus of the candidate object is constructed.

For example, the position coordinates of all the sensors acquiring the candidate object in the control circle may be sorted according to the time sequence of the candidate object passing through the corresponding sensors, and then the sorted sensors are connected to obtain the movement track of the candidate object. As shown in fig. 4, the moving trajectory of the candidate objects is shown, wherein the moving trajectory of the first candidate object is from the

sensor

01, 02, 03, 04 to the sensor 05, the moving trajectory of the second candidate object is from the sensor 01 to the sensor 06, the moving trajectory of the third candidate object is from the sensor 06 to the sensor 05, and the moving trajectory of the fourth candidate object is from the sensor 06, the sensor 07 to the sensor 05.

In step 302, the movement trajectory is input into the neural network model, and the confidence of the next movement trajectory of the candidate object output by the neural network model is obtained, and the confidence is used for representing the probability that the next movement trajectory is the movement trajectory of the normal object.

In a possible implementation manner, in order to ensure the accuracy of the output result of the neural network model, the neural network model used in the embodiment of the present application first needs to be trained, and specifically, the following steps as shown in fig. 5 may be performed:

in step 501, a training sample is obtained, where the training sample includes a historical movement trajectory of a sample object, a next movement trajectory of the historical movement trajectory, and a target confidence.

In step 502, the historical movement trajectory of the sample object is input into the neural network model, and the next movement trajectory and the prediction confidence of the sample object output by the neural network model are obtained.

In step 503, the next movement trajectory output by the neural network model is compared with the next movement trajectory included in the training sample to obtain a position loss, and the confidence loss is determined based on the prediction confidence and the target confidence.

In step 504, the neural network model is trained using the position loss and the confidence loss.

Illustratively, the next-step movement track and the prediction confidence of the predicted historical movement track can be output by inputting the historical movement track of the sample object into the neural network model, then the position loss of the two next-step movement tracks can be obtained by comparing the next-step movement track of the sample object with the prediction confidence, the confidence loss can be determined by comparing the prediction confidence with the target confidence, so that the parameters in the neural network model can be adjusted according to the position loss and the confidence loss, the parameters of the neural network model can be more stable by training the neural network model by using a plurality of sample objects, and the confidence of the next-step movement track of the candidate object can be obtained by inputting the movement track of the candidate object into the trained neural network model.

Exemplarily, as shown in fig. 6, a schematic diagram of the prediction of the movement trajectory of the candidate object is shown. Wherein, the sensors of different candidate objects appearing in the next step have certain rules, the circles represent the sensors, the solid arrows represent the movement tracks of the candidate objects, and the dotted lines represent the movement tracks predicted in the next step. Abstracting the moving track of the candidate object represented by the solid arrow into a track point sequence{t₁，t₂，.....t_nAnd then inputting a time series prediction model based on a transform (neural network model), wherein the movement track of the candidate object at the next moment represented by a dotted line and the confidence of the next movement track of the candidate object can be predicted.

In step 303, it is determined that the confidence of the candidate object is less than a preset confidence threshold.

In a possible implementation manner, the risk coefficient of the candidate object obtained according to the step shown in fig. 2 and the confidence coefficient of the next movement trajectory of the candidate object obtained according to the step shown in fig. 3 are mutually verified, so as to determine whether the detection result of the candidate object is an abnormal object, and then the abnormal object early warning model performs early warning based on the detection result of the abnormal object. All candidate objects in the control circle are detected based on the method provided by the application to obtain corresponding detection results, the abnormal object early warning model carries out early warning based on the abnormal object detection results, and finally early warning of all abnormal objects in the control circle is completed. For example, if the risk coefficient of the candidate is greater than or equal to the preset risk coefficient threshold and the confidence of the candidate is less than the preset confidence threshold, determining that the candidate is an abnormal object. Therefore, although the candidate object can be determined as the abnormal object according to the condition that the risk coefficient of the candidate object is greater than or equal to the preset risk coefficient threshold or the confidence coefficient of the candidate object is smaller than the preset confidence coefficient threshold, the detection result can be more accurate through mutual verification of the candidate object and the candidate object, and the accuracy and the real-time effectiveness of the early warning of the abnormal personnel can be improved.

Based on the foregoing description, the present application discloses a method for detecting an abnormal object, which summarizes various dimensional characteristics of the abnormal object by establishing a prior probability table in advance. And after the abnormal object invades, a sensor is triggered to alarm, and then data collected by each sensor is searched in a prevention and control circle taking the sensor as a reference. Then, the characteristics of the abnormal object are extracted from the collected data based on the prior probability table, the prior probability corresponding to the extracted characteristics is searched from the prior probability table, and then the extracted characteristics and the prior probability are processed by adopting a Bayesian probability model, so that which objects in the collected data are the alarming abnormal objects can be predicted. Therefore, the multi-dimensional characteristics of the candidate object are automatically extracted, and the prior probability corresponding to the characteristics is input into the Bayesian probability model by utilizing the algorithm, so that the abnormal object can be detected, the abnormal object is early warned, the real-time effectiveness and accuracy of early warning are further ensured, and the difficulty of manual checking is reduced.

In addition, the method and the device further acquire the historical moving track and the next moving track of the sample object, train the neural network model to obtain the confidence coefficient of the next moving track output by the neural network model and the next moving track in the training sample, predict the confidence coefficient of the next moving track of the candidate object through the trained neural network model according to the historical moving track of the candidate object, and mutually verify the confidence coefficient of the next moving track of the candidate object, so that the early warning accuracy of abnormal personnel is further ensured. By using the learning training model, updating and adjusting are continuously carried out according to the early warning result, so that the model is continuously optimized, the more the model is used, and abnormal personnel can be accurately early warned in real time.

As shown in fig. 7, based on the same inventive concept as the above-mentioned abnormal object detection method, an embodiment of the present application further provides an abnormal object detection apparatus, including: an alarm position obtaining module 701, a data obtaining module 702, a multidimensional feature extraction module 703, a prior probability obtaining module 704, a risk coefficient determining module 705, and an anomaly identification module 706, wherein:

an alarm position obtaining module 701, configured to obtain alarm information of an abnormal object, where the alarm information includes a sensor position that sends the alarm information;

a data obtaining module 702, configured to obtain data collected by a sensor to be processed around the sensor position by using the sensor position as a reference;

a multidimensional feature extraction module 703, configured to extract feature information of a candidate object from data acquired by a sensor to be processed, where the feature information includes multidimensional features;

a prior probability obtaining module 704, configured to obtain a prior probability corresponding to a feature of each dimension in the multi-dimensional features from a preset prior probability table; wherein the prior probabilities include a prior probability that a candidate is the anomalous object and a prior probability that the candidate is not the anomalous object;

a risk coefficient determining module 705, configured to process the multidimensional features and the prior probability corresponding to the feature of each dimension by using a bayesian probability model, to obtain a risk coefficient of the candidate object; wherein the risk factor is indicative of a probability that the candidate object is the abnormal object;

an anomaly identification module 706, configured to determine that the candidate object is the abnormal object if the risk coefficient of the candidate object is greater than or equal to a preset risk coefficient threshold.

the anomaly identification module 706 is specifically configured to determine that the confidence level of the candidate object is smaller than a preset confidence level threshold.

In a possible implementation, the multidimensional features extracted by the multidimensional feature extraction module 703 include:

In a possible implementation manner, before the prior probability obtaining module 704 is configured to obtain the prior probability corresponding to the feature of each dimension in the multi-dimensional feature from a preset prior probability table, the apparatus further includes:

acquiring data of the candidate objects acquired by a plurality of sensors;

In one possible embodiment, the apparatus comprises:

The detection device for the abnormal object and the detection method for the abnormal object provided by the embodiment of the application adopt the same inventive concept, can obtain the same beneficial effects, and are not repeated herein.

Based on the same inventive concept as the detection method of the abnormal object, the embodiment of the application also provides the electronic equipment. An electronic device 800 according to this embodiment of the application is described below with reference to fig. 8. The electronic device 800 shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

As shown in fig. 8, the electronic device 800 is represented in the form of a general electronic device. The components of the electronic device 800 may include, but are not limited to: the at least one processor 801, the at least one memory 802, and a bus 803 that couples various system components including the memory 802 and the processor 801.

Bus 803 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a processor, or a local bus using any of a variety of bus architectures.

The memory 802 may include readable media in the form of volatile memory, such as Random Access Memory (RAM)8021 and/or cache memory 8022, and may further include Read Only Memory (ROM) 8023.

Memory 802 may also include a program/utility 8025 having a set (at least one) of program modules 8024, such program modules 8024 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

The electronic device 800 may also communicate with one or more external devices 804 (e.g., keyboard, pointing device, etc.), with one or more devices that enable a user to interact with the electronic device 800, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 800 to communicate with one or more other electronic devices. Such communication may be through input/output (I/O) interfaces 805. Also, the electronic device 800 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the internet) via the network adapter 806. As shown, the network adapter 806 communicates with other modules for the electronic device 800 over the bus 803. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

In an exemplary embodiment, a computer-readable storage medium comprising instructions, such as memory 802 comprising instructions, executable by processor 801 to perform the above-described method of detecting an anomalous object is also provided. Alternatively, the storage medium may be a non-transitory computer readable storage medium, for example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product is also provided, comprising a computer program which, when being executed by the processor 801, carries out any one of the methods of detecting an anomalous object as provided herein.

In an exemplary embodiment, various aspects of a method for detecting an abnormal object provided by the present application may also be implemented in the form of a program product including program code for causing a computer device to perform the steps in the method for detecting an abnormal object according to various exemplary embodiments of the present application described above in this specification when the program product is run on the computer device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The program product of the detection method for an abnormal object of the embodiments of the present application may employ a portable compact disc read only memory (CD-ROM) and include program codes, and may be run on an electronic device. However, the program product of the present application is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the consumer electronic device, partly on the consumer electronic device, as a stand-alone software package, partly on the consumer electronic device and partly on a remote electronic device, or entirely on the remote electronic device or server. In the case of remote electronic devices, the remote electronic devices may be connected to the consumer electronic device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external electronic device (e.g., through the internet using an internet service provider).

It should be noted that although several units or sub-units of the apparatus are mentioned in the above detailed description, such division is merely exemplary and not mandatory. Indeed, the features and functions of two or more units described above may be embodied in one unit, according to embodiments of the application. Conversely, the features and functions of one unit described above may be further divided into embodiments by a plurality of units.

Further, while the operations of the methods of the present application are depicted in the drawings in a particular order, this does not require or imply that these operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable image scaling apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable image scaling apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable image scaling apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable image scaling device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer implemented process such that the instructions which execute on the computer or other programmable device provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.

Claims

1. A method for detecting an abnormal object, the method comprising:

2. The method of claim 1, wherein if it is determined that the risk factor of the candidate object is greater than or equal to a predetermined risk factor threshold and before it is determined that the candidate object is the abnormal object, the method further comprises:

3. The method of claim 1, wherein the multi-dimensional features comprise:

4. The method according to claim 1, wherein before obtaining the prior probability corresponding to the feature of each dimension in the multi-dimensional feature from the preset prior probability table, the method further comprises:

5. The method according to claim 2, wherein the extracting the movement trajectory of the candidate object from the data collected by the plurality of sensors specifically comprises:

acquiring data of the candidate objects acquired by a plurality of sensors;

6. The method of claim 2, wherein training the neural network model comprises:

7. The method according to any one of claims 1-6, further comprising:

8. An apparatus for detecting an abnormal object, the apparatus comprising:

9. An electronic device, comprising:

a processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the method of detecting an anomalous object as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, wherein instructions in the computer-readable storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the method of detecting an abnormal object according to any one of claims 1 to 7.