Mobile checking and recording system

Abstract

The invention discloses a mobile checking and recording system, which relates to the technical field of target object monitoring, and is characterized in that under the condition that fingerprints of monitoring points do not meet requirements, a target object feature extraction component performs feature extraction on portrait information so as to judge the state of a target object; the target object can be monitored in a whole process by integrating multi-aspect data, and the target object monitoring efficiency and precision are improved; the abnormal monitoring point position is identified through the monitoring point fingerprint, and whether deep detection is carried out or not is determined by combining the change characteristics of the monitoring point position in unit time; feature extraction is carried out on the face contour data and the figure data, feature extraction results of the two kinds of data are combined to judge the state of the target object, and the target object monitoring precision is improved.

Description

Technical Field

[0001] The present invention belongs to the field of target object monitoring, and relates to a target object monitoring technology based on multi-source data, and specifically is a mobile verification and recording system. Background Art

[0002] With the continuous improvement of the level of informatization, big data and machine learning technologies are gradually applied to relevant research in the field. In recent years, the use of machine learning technology to identify the target objects involved has also appeared in applications, mainly divided into regional identification, specific types of target object integral monitoring models, etc.; the specific type of target object integral monitoring model mainly uses artificial experience combined with data statistics and other methods, combined with various business data of historical target objects of specific cases, to establish the target object's integral information, so as to monitor the target object.

[0003] The existing point monitoring for specific types of target objects is mainly based on the office experience of relevant personnel, rather than on the research on the correlation between abnormal behavior information and the occurrence of cases. The existing points rules, related indexes, specific scores, points coefficients, etc. lack scientificity and the basis of empirical research. The method of setting scores through manual experience or statistical data lacks scientific basis, resulting in unclear monitoring effects. When monitoring the health of target objects, a large amount of data is generated every day. Therefore, in the face of massive monitoring data, it is necessary to use reliable algorithms to extract valuable information from the features, thereby providing important scientific reference for the status assessment of the target objects.

[0004] In the process of monitoring target objects, the existing technology considers fewer influencing factors, resulting in low target object monitoring accuracy and a small scope of application; therefore, there is an urgent need for a mobile verification and recording system. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a mobile verification and recording system for solving the technical problems that the prior art considers fewer influencing factors during target object monitoring, resulting in low target object monitoring accuracy and a small scope of application.

[0006] The present invention obtains the monitoring point fingerprint corresponding to the monitoring point coordinates through a fingerprint collection device connected to a target object information integration component. When the monitoring point fingerprint does not meet the requirements, the target object feature extraction component extracts features from the portrait information to determine the state of the target object. It can integrate multiple aspects of data to monitor the target object throughout the entire process, thereby improving the efficiency and accuracy of target object monitoring.

[0007] To achieve the above-mentioned object, a first aspect of the present invention provides a mobile verification and recording system, comprising a target object feature extraction component and a target object information integration component connected thereto;

[0008] Target object information integration component: collects the monitoring point fingerprint corresponding to the monitoring point coordinates according to the fingerprint collection device; and collects the portrait information of the target object according to the extended instruction test; wherein the portrait information includes facial contour data and body shape data;

[0009] Target object feature extraction component: determine the monitoring point position according to the target object recognition algorithm, convert the monitoring point position into the monitoring point coordinates; feature extract the monitoring point fingerprint, and generate the extended instruction; and

[0010] The facial contour directional gradient histogram algorithm or the Haar-like feature model is combined to perform feature extraction on the portrait information, and the state of the target object is determined according to the feature extraction result.

[0011] Preferably, the target object feature extraction component is connected to a plurality of the target object information integration components;

[0012] The target object information integration component is connected to the fingerprint collection device and various types of sensors; wherein the fingerprint collection device includes a fingerprint collector and a carrier device thereof.

[0013] Preferably, the target object feature extraction component obtains the monitoring point coordinates according to a target object recognition algorithm, including:

[0014] Establishing the target object recognition algorithm according to the target object design data of the target object;

[0015] According to the target object recognition algorithm, conventional detection points of the target object are selected and marked as the monitoring point positions; thereafter, the monitoring point positions are converted into the monitoring point coordinates and sent to the target object information integration component.

[0016] Preferably, the target object information integration component obtains the monitoring point coordinates to obtain the corresponding monitoring point fingerprint, including:

[0017] The target object information integration component extracts features from the received monitoring point coordinates, and calls the fingerprint collection state to collect the corresponding conventional fingerprint according to the feature extraction result;

[0018] The conventional fingerprint is verified and then sent to the target object feature extraction component.

[0019] Preferably, the target object feature extraction component extracts features from the conventional fingerprint and generates the extended instruction according to the feature extraction result, including:

[0020] Identifying anomalies in the positions of monitoring points in the conventional fingerprint by a fingerprint recognition algorithm;

[0021] The feature extraction is to extract the variation characteristics of the monitoring point position per unit time, and the extension instruction is automatically generated according to the feature extraction result.

[0022] Preferably, the target object feature extraction component extracts features from the facial contour data in combination with the facial contour directional gradient histogram algorithm to obtain facial contour feature extraction time, including:

[0023] Extracting features from the received facial contour data, and establishing a facial contour change curve with time as an independent variable; wherein the facial contour data includes facial features data and bone data;

[0024] Calling the corresponding facial contour directional gradient histogram algorithm according to the facial contour data;

[0025] The facial contour change curve is compared with the corresponding facial contour directional gradient histogram algorithm, and the facial contour feature extraction time is generated and set according to the comparison result.

[0026] Preferably, the target object feature extraction component extracts features from the body shape data in combination with the Haar-like feature model to obtain the body shape feature extraction time, including:

[0027] Receive the body shape data, and convert the body shape data to obtain a body shape code; wherein the body shape data includes height data of the target object, limb data of the target object, disability data of the target object and corresponding weight data;

[0028] Calling the Haar-like feature model; wherein the Haar-like feature model is established based on a deep reinforcement algorithm;

[0029] The body shape code is input into the Haar-like feature model to obtain the corresponding body shape feature extraction time.

[0030] Preferably, establishing the Haar-like feature model based on the deep enhancement algorithm includes:

[0031] Update and obtain database target object data; wherein the database target object data is obtained in a daily environment, including body shape change data and corresponding body shape change time;

[0032] Constructing the deep reinforcement algorithm; wherein the deep reinforcement algorithm includes the DQN algorithm or the DUELING-DQN algorithm;

[0033] The deep reinforcement algorithm is trained using the database target object data, and the trained deep reinforcement algorithm is marked as the Haar-like feature model.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. The present invention obtains the monitoring point fingerprint corresponding to the monitoring point coordinates through a fingerprint collection device connected to the target object information integration component. When the monitoring point fingerprint does not meet the requirements, the target object feature extraction component extracts features from the portrait information to determine the state of the target object. It can integrate multiple aspects of data to monitor the target object throughout the entire process, thereby improving the efficiency and accuracy of target object monitoring.

[0036] 2. The present invention identifies abnormal monitoring point positions through monitoring point fingerprints, and determines whether to conduct in-depth detection based on the change characteristics of the monitoring point positions per unit time; extracts features from facial contour data and body shape data respectively, and combines the feature extraction results of the two types of data to determine the state of the target object, using multiple data as the basis for feature extraction to improve the monitoring accuracy of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 It is a schematic diagram of the system components of the present invention;

[0039] Figure 2 A first flow chart for implementing the system of the present invention;

[0040] Figure 3 A second flow chart is implemented for the system of the present invention;

[0041] Figure 4 A third flow chart for implementing the system of the present invention; DETAILED DESCRIPTION

[0042] The technical scheme of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The present invention obtains the monitoring point fingerprint corresponding to the monitoring point coordinates through a fingerprint collection device connected to a target object information integration component. When the monitoring point fingerprint does not meet the requirements, the target object feature extraction component extracts features from the portrait information to determine the state of the target object. It can integrate multiple aspects of data to monitor the target object throughout the entire process, thereby improving the efficiency and accuracy of target object monitoring.

[0044] like Figure 1 As shown, the first embodiment of the present application provides a mobile verification and recording system, including a target object feature extraction component and a target object information integration component connected thereto;

[0045] Target object information integration component: collects the monitoring point fingerprint corresponding to the monitoring point coordinates according to the fingerprint collection device; and collects the portrait information of the target object according to the extended instruction test;

[0046] Target object feature extraction component: Determine the monitoring point location based on the target object recognition algorithm and convert the monitoring point location into monitoring point coordinates; feature extract the monitoring point fingerprint and generate extended instructions; and extract the features of the portrait information in combination with the facial contour directional gradient histogram algorithm or the Haar-like feature model, and determine the state of the target object based on the feature extraction results.

[0047] In this application, the target object feature extraction component is equivalent to a central processing unit, which can process a variety of data in a timely manner, and at the same time, establish various models based on multi-source historical data, responsible for large-scale data processing. The target object information integration component is equivalent to a data transfer station with data processing capabilities, responsible for collecting data and performing data preprocessing.

[0048] In the present application, the target object feature extraction component is connected to several target object information integration components; the target object information integration component is connected to the fingerprint collection device and various types of sensors; the fingerprint collection device includes a fingerprint collector and its carrier device.

[0049] The fingerprint collection device is mainly used to collect fingerprint data at regular locations. Considering that the regular detection cycle of the target object is relatively long (routine detection is performed approximately once every two years), the fingerprint collector can be fixed on the fingerprint collector and the fingerprint collector can be controlled to capture the coordinates of the monitoring point to obtain the corresponding fingerprint data. The fingerprint collector here can be understood as the above-mentioned carrier device.

[0050] Various types of sensors are used to collect weight data and some portrait information. Various types of sensors collect data according to the set collection cycle and send the data to the target object information integration component in a timely manner.

[0051] In this application, the target object feature extraction component obtains the coordinates of the monitoring point according to the target object recognition algorithm, including:

[0052] establishing a target object recognition algorithm based on target object design data of the target object;

[0053] According to the target object recognition algorithm, regular detection points of the target object are selected and marked as monitoring point positions; then, the monitoring point positions are converted into monitoring point coordinates and sent to the target object information integration component.

[0054] Before detecting the target object, it is necessary to determine which positions need to be focused on. It is not necessary to detect all positions of the target object. It is sufficient to detect the positions that need to be focused on, which can reduce redundant data and detection costs.

[0055] The target object feature extraction component first establishes a target object recognition algorithm based on the target object design data, and then delineates regular detection points based on the target object recognition algorithm, so as to obtain the corresponding regular positions. It should be noted that the regular detection points can be manually selected by experienced personnel or selected by a machine learning algorithm according to pre-set rules.

[0056] Conventional detection points or conventional locations are generally locations that need to be noted in the routine detection of target objects. After determining the conventional locations, they need to be converted into monitoring point coordinates based on geographic data. The fingerprint collection device collects fingerprint data based on the monitoring point coordinates. It is understandable that each target object should include several monitoring point coordinates.

[0057] In this application, the target object information integration component obtains the monitoring point coordinates to obtain the corresponding monitoring point fingerprint, including:

[0058] The target object information integration component extracts features from the received monitoring point coordinates, and calls the fingerprint collection state to collect the corresponding conventional fingerprint according to the feature extraction result; the conventional fingerprint is verified and sent to the target object feature extraction component.

[0059] Acquiring the monitoring point fingerprint at the monitoring point coordinates is the first step in extracting the target object's features and is also the basis for routine detection of the target object. After receiving the monitoring point coordinates, the fingerprint collector (fingerprint collection device) takes a picture of the monitoring point position corresponding to the monitoring point coordinates to obtain the corresponding monitoring point fingerprint. The fingerprint collection task of the fingerprint collector is completed only after all the monitoring point fingerprints corresponding to the monitoring point coordinates are collected and qualified.

[0060] It should be noted that after the fingerprint collector collects the fingerprint of the monitoring point corresponding to a certain conventional coordinate, it will be sent to the target object feature extraction component immediately. The target object feature extraction component verifies the conventional fingerprint, such as position accuracy, clarity, etc. After the verification, the fingerprint preprocessing (fingerprint cropping, grayscale conversion, etc.) is performed and sent to the target object feature extraction component. If the conventional fingerprint fails the verification, a command is sent to the fingerprint collector to re-collect the conventional fingerprint.

[0061] like Figure 2 As shown, in this application, the target object feature extraction component extracts features from conventional fingerprints and generates extended instructions based on the feature extraction results, including:

[0062] The fingerprint recognition algorithm is used to identify the abnormality of the monitoring point position in the conventional fingerprint; the change characteristics of the abnormal monitoring point position per unit time are extracted, and the extended instruction is automatically generated according to the feature extraction result.

[0063] After receiving the conventional fingerprint of the target object, the target object feature extraction component extracts features to determine whether there is an abnormality in the conventional position, thereby completing the conventional detection of the target object. The construction of different conventional positions corresponds to different abnormalities.

[0064] After identifying the anomaly, it is necessary to extract features of the number and distribution of anomalies to determine whether further detection is required, that is, whether it is necessary to generate and send extended instructions. For example:

[0065] Determine whether to generate and send an extended instruction based on the number of abnormalities in regular positions; count the number of abnormalities in regular positions in the target object; when the number of abnormalities exceeds a quantity threshold, generate and send an extended instruction; wherein the quantity threshold is flexibly set in combination with the target object type, service life, etc.

[0066] Determine whether to generate a send extension instruction based on the distribution of regular positions. Count the number of abnormalities in regular positions in the target object; when the number of abnormalities does not exceed the threshold, select an abnormal regular position, extract features to see if there are a certain number of abnormal regular positions around it, and generate a send extension instruction if there are.

[0067] It should be noted that when the target object feature extraction component does not generate an extended instruction, the abnormal monitoring point positions are sent to personnel, and the personnel process these abnormal monitoring point positions.

[0068] like Figure 3 As shown, in the present application, the target object feature extraction component combines the facial contour directional gradient histogram algorithm to extract features from the facial contour data, and obtains the facial contour feature extraction time, including:

[0069] Extract features from the received facial contour data, and establish a facial contour change curve with time as an independent variable;

[0070] Call the corresponding facial contour directional gradient histogram algorithm according to the facial contour data;

[0071] The facial contour change curve is compared with the corresponding facial contour directional gradient histogram algorithm, and the facial contour feature extraction time is set according to the comparison result.

[0072] Facial contour data includes facial features data, bone data, etc., and a facial contour change curve is established with time as the independent variable and facial contour data as the dependent variable. The facial contour change curve is compared with the facial contour directional gradient histogram algorithm to determine whether the facial contour data is abnormal. The facial contour change curve includes facial features data curve, bone data curve, etc.

[0073] The facial contour directional gradient histogram algorithm is obtained in daily environments, and the facial contour directional gradient histogram algorithm for multiple scenes should be established according to the environment and age of the test target object. The facial contour directional gradient histogram algorithm is stored in the target object feature extraction component and is updated regularly.

[0074] When the similarity between the facial contour change curve and the corresponding facial contour directional gradient histogram algorithm is low, the generated facial contour feature extraction time is set to A; otherwise, the facial contour feature extraction time is not generated, or the generated facial contour feature extraction time is set to B.

[0075] In this application, the target object feature extraction component combines the Haar-like feature model to extract features from the body shape data and obtain the body shape feature extraction time, including:

[0076] Receiving body shape data, and converting the body shape data to obtain a body shape code;

[0077] Call the Haar-like feature model, input the body shape code into the Haar-like feature model, and obtain the corresponding body shape feature extraction time.

[0078] The body shape data includes the target object's height data, target object's limb data, target object's disability data and corresponding weight data. Taking into account the nonlinear properties of the body shape data, the body shape data is processed in combination with a deep enhancement algorithm with powerful nonlinear processing capabilities. This can simplify operations and quickly determine whether the body shape data is abnormal.

[0079] like Figure 3 As shown, in an optional embodiment, a Haar-like feature model is established based on a deep reinforcement algorithm, including:

[0080] Update and obtain the target object data in the database and build a deep reinforcement algorithm;

[0081] The deep reinforcement algorithm is trained through the target object data in the database, and the trained deep reinforcement algorithm is marked as a Haar-like feature model.

[0082] The database target object data is obtained in daily environment, including body shape change data and corresponding body shape change time; the body shape change data is consistent with the content attribute of the above body shape data, that is, it includes the same data type, but the data value may be different. It should be noted that the body shape change time corresponding to the body shape change data is set by professional personnel to ensure its accuracy.

[0083] The target object feature extraction component determines the state of the target object according to the facial contour feature extraction time and the body shape feature extraction time, locates the abnormal position according to the body shape feature extraction time or the body shape feature extraction time, and dispatches personnel to inspect and maintain the abnormal position. So far, this application has completed the routine detection of the target object and further in-depth detection, realized the full range of detection, and helped to ensure the safety of the target object.

[0084] Working principle of the present invention:

[0085] The target object information integration component collects the monitoring point fingerprint corresponding to the monitoring point coordinates according to the fingerprint collection device; and collects the portrait information of the target object according to the extended instruction test.

[0086] The target object feature extraction component extracts the fingerprint of the monitoring point and generates an extended instruction; and extracts the features of the portrait information in combination with the facial contour directional gradient histogram algorithm or the Haar-like feature model, and determines the state of the target object according to the feature extraction result.

[0087] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A mobile verification and recording system, comprising a target object feature extraction component and a target object information integration component connected thereto, characterized in that: Target object information integration component: collects the monitoring point fingerprint corresponding to the monitoring point coordinates according to the fingerprint collection device; and collects the portrait information of the target object according to the extended instruction test; wherein the portrait information includes facial contour data and body shape data; Target object feature extraction component: determine the monitoring point position according to the target object recognition algorithm, convert the monitoring point position into the monitoring point coordinates; feature extract the monitoring point fingerprint and generate the extended instruction; and extract the features of the portrait information in combination with the facial contour directional gradient histogram algorithm or the Haar-like feature model, and determine the state of the target object according to the feature extraction result.

2. A mobile verification and recording system according to claim 1, characterized in that: The target object feature extraction component is connected to a plurality of target object information integration components; The target object information integration component is connected to the fingerprint collection device and various types of sensors; wherein the fingerprint collection device includes a fingerprint collector and a carrier device thereof.

3. A mobile verification and recording system according to claim 1, characterized in that: The target object feature extraction component obtains the monitoring point coordinates according to the target object recognition algorithm, including: Establishing the target object recognition algorithm according to the target object design data of the target object; According to the target object recognition algorithm, conventional detection points of the target object are selected and marked as the monitoring point positions; thereafter, the monitoring point positions are converted into the monitoring point coordinates and sent to the target object information integration component.

4. A mobile verification and recording system according to claim 3, characterized in that: The target object information integration component obtains the monitoring point coordinates to obtain the corresponding monitoring point fingerprint, including: The target object information integration component extracts features from the received monitoring point coordinates, and calls the fingerprint collection state to collect the corresponding conventional fingerprint according to the feature extraction result; The conventional fingerprint is verified and then sent to the target object feature extraction component.

5. A mobile verification and recording system according to claim 4, characterized in that: The target object feature extraction component extracts features from the conventional fingerprint and generates the extended instruction according to the feature extraction result, including: Identifying anomalies in the positions of monitoring points in the conventional fingerprint by a fingerprint recognition algorithm; The feature extraction is to extract the variation characteristics of the monitoring point position per unit time, and the extension instruction is automatically generated according to the feature extraction result.

6. A mobile verification and recording system according to claim 1, characterized in that: The target object feature extraction component extracts features from the facial contour data in combination with the facial contour directional gradient histogram algorithm to obtain facial contour feature extraction time, including: Extracting features from the received facial contour data, and establishing a facial contour change curve with time as an independent variable; wherein the facial contour data includes facial features data and bone data; Calling the corresponding facial contour directional gradient histogram algorithm according to the facial contour data; The facial contour change curve is compared with the corresponding facial contour directional gradient histogram algorithm, and the facial contour feature extraction time is generated and set according to the comparison result.

7. A mobile verification and recording system according to claim 1, characterized in that: The target object feature extraction component extracts features from the body shape data in combination with the Haar-like feature model to obtain a body shape feature extraction time, including: Receive the body shape data, and convert the body shape data to obtain a body shape code; wherein the body shape data includes height data of the target object, limb data of the target object, disability data of the target object and corresponding weight data; Calling the Haar-like feature model; wherein the Haar-like feature model is established based on a deep reinforcement algorithm; The body shape code is input into the Haar-like feature model to obtain the corresponding body shape feature extraction time.

8. A mobile verification and recording system according to claim 7, characterized in that: The Haar-like feature model is established based on the deep enhancement algorithm, including: Update and obtain database target object data; wherein the database target object data is obtained in a daily environment, including body shape change data and corresponding body shape change time; Constructing the deep reinforcement algorithm; wherein the deep reinforcement algorithm includes the DQN algorithm or the DUELING-DQN algorithm; The deep reinforcement algorithm is trained using the database target object data, and the trained deep reinforcement algorithm is marked as the Haar-like feature model.

Images