CN115409992A - Remote driving patrol car system - Google Patents
Remote driving patrol car system Download PDFInfo
- Publication number
- CN115409992A CN115409992A CN202210967369.4A CN202210967369A CN115409992A CN 115409992 A CN115409992 A CN 115409992A CN 202210967369 A CN202210967369 A CN 202210967369A CN 115409992 A CN115409992 A CN 115409992A
- Authority
- CN
- China
- Prior art keywords
- data
- module
- patrol car
- group
- driving
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 claims abstract description 78
- 238000004891 communication Methods 0.000 claims abstract description 24
- 238000000034 method Methods 0.000 claims abstract description 22
- 238000004458 analytical method Methods 0.000 claims abstract description 14
- 238000011084 recovery Methods 0.000 claims description 44
- 238000007689 inspection Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000012544 monitoring process Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 230000007774 longterm Effects 0.000 claims description 3
- 238000011897 real-time detection Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/0223—User address space allocation, e.g. contiguous or non contiguous base addressing
- G06F12/023—Free address space management
- G06F12/0253—Garbage collection, i.e. reclamation of unreferenced memory
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses a remote driving patrol car system, which belongs to the field of unmanned driving and comprises a control platform, a vehicle-mounted camera, a communication connection module, a target detection module, a parameter adjustment module, a fault detection module, a built-in storage module, a cloud database and an alarm module, wherein the vehicle-mounted camera is arranged on the control platform; according to the method, the collected image data is subjected to cascade analysis by constructing the primary target detection network and the secondary target detection network, so that the information of each object in a patrol environment can be more accurately analyzed, the accuracy of target detection of the patrol car is improved, the analysis efficiency of related personnel is improved, the workload of the related personnel is reduced, the data stored in the system can be automatically recovered and deleted, the memory utilization rate of the system is improved, the communication transmission efficiency is prevented from being reduced due to excessive redundant data, and the stable operation of the system is ensured.
Description
Technical Field
The invention relates to the field of unmanned driving, in particular to a remote driving patrol car system.
Background
Remote control means that a manager dials in a different place through a computer network or both access the Internet and the like to communicate equipment to be controlled, acquired data of the controlled equipment is displayed on a computer of the manager, the remote equipment is configured, software is installed, modified and the like through a local computer, and the remote control becomes practical along with the appearance of a parallel driving technology based on 5G communication, different from remote control, the parallel driving control has the characteristics of long control distance, small time delay and high reliability, so that the technology is gradually applied to patrol cars to facilitate daily patrol of related personnel and effectively save manpower;
through retrieval, chinese patent No. CN202111466122.6 discloses a remote driving control system and method for a crawler-type patrol car, and the system and method eliminate the shake caused by the operation of a driver, optimize the mapping relation between the pedal travel and the speed and improve the driving experience, but the accuracy of target detection is low, the analysis efficiency of related personnel is reduced, and the workload of the related personnel is increased; in addition, the existing remote driving patrol car system is low in memory utilization rate, low in communication transmission efficiency due to excessive redundant data and poor in system operation stability, and therefore a remote driving patrol car system is provided.
Disclosure of Invention
The invention aims to solve the defects in the prior art and provides a remote driving patrol car system.
In order to achieve the purpose, the invention adopts the following technical scheme:
a remote driving patrol car system comprises a control platform, a vehicle-mounted camera, a communication connection module, a target detection module, a parameter adjustment module, a fault detection module, a built-in storage module, a cloud database and an alarm module;
the control platform is used for a manager to log in and control the patrol car to run, and meanwhile, feedback information of each sub-module is received and is provided for the manager to check in a graphical mode;
the vehicle-mounted camera is used for acquiring surrounding environment information of the patrol car to generate image data;
the communication connection module is used for receiving a control instruction sent by the control platform and carrying out communication detection;
the target detection module is used for receiving the image data and analyzing and feeding back the image data;
the parameter adjusting module is used for collecting the driving parameters of the patrol car and then carrying out optimization adjustment on the patrol car;
the fault detection module is used for receiving the driving data of the patrol car at any time and carrying out fault feedback;
the built-in storage module is used for temporarily storing all groups of data of the patrol car and periodically recovering the data;
the cloud database is used for receiving and storing each group of data uploaded by the control platform;
the alarm module is used for receiving each group of data fed back by the target detection module and sending alarm information to a manager according to the data content.
As a further scheme of the present invention, the communication connection module specifically comprises the following steps:
the method comprises the following steps: the communication connection module creates a group of sockets in the form of file descriptors and binds the generated sockets with a local IP address port to set monitoring;
step two: after the monitoring setting is successful, the communication connection module waits for and receives a connection request of a client, acquires a group of new file descriptors and calls an accept function to detect a read buffer area of the file descriptors;
step three: if the read buffer area can not detect the data, the function is blocked and the connection is disconnected, and if the read buffer area detects the data, the function is unblocked and the connection is established.
As a further scheme of the present invention, the specific steps of analyzing and feeding back by the target detection module are as follows:
step (1): the target detection module processes video or image sequence frames with fixed frame rate in the image data, then calculates and records the interval time of actual video frames, simultaneously constructs a primary target detection network to extract inspection pictures from the image data frame by frame, and constructs a picture data set according to different sizes of each group of inspection pictures;
step (2): then, zooming each group of acquired environment inspection pictures according to different resolutions set by an administrator, constructing an image pyramid according to the zoomed inspection pictures, extracting feature information of each group of inspection pictures in the picture data set, sending each group of feature information into a bidirectional feature pyramid for feature fusion, carrying out scale normalization processing on the feature information by the image pyramid, and outputting a detection frame, a category and a score;
and (3): collecting object detection frame information in the inspection pictures, generating corresponding detection frame coordinates, performing expanded cutting on related inspection pictures, collecting each group of enlarged and cut object pictures, performing temporary storage through a built-in storage module, constructing a secondary target detection network, filtering out simple negative samples belonging to a background in each group of object pictures, and selecting areas possibly containing objects for classification and regression;
and (4): nine anchor frames with specified proportion and size are generated at each point of the object pictures with high or low semantic information, the anchor frames are classified and regressed, the object positions in each group of object pictures are detected through enlarged clipping, the detection results are fed back to the control platform, and target tracking is carried out at the same time.
As a further scheme of the present invention, the specific steps of the target tracking in step (4) are as follows:
the first step is as follows: establishing a motion model through a Kalman filtering theory, simultaneously acquiring the motion state of each object in real time through the established motion model, and then defining the motion state of each object in a video frame by the motion model according to the linear motion assumption of each object;
the second step is that: collecting the motion state of each object in the current video frame, constructing a prediction equation to estimate the motion state of each object in the next video frame, and calculating the detection results of all objects in the current video frame of the ith vehicle-mounted camera according to a multi-target real-time detection algorithm;
the third step: calculating the cosine distance between the detection result and the object, filtering through a manually set threshold value to obtain a cost matrix represented by the cosine distance, and then performing binary matching to obtain a matching result set;
the fourth step: and the target detection module updates Kalman gain and covariance matrixes in the motion model of the object which finally fails to be matched, matches each group of objects which finally fails to be matched again after the updating is finished, processes video frame data of each path of video stream in parallel after the matching is finished, and sequentially executes object marking, estimation of the motion state of the object, matching correlation and cross-monitoring multi-object tracking on video frames obtained after down-sampling in each path of video stream.
As a further scheme of the present invention, the parameter adjusting module comprises the following specific steps:
s1.1: the method comprises the following steps that a parameter adjusting module receives all driving data of a patrol car, an optimized neural network corresponding to the patrol car is constructed, and the optimized neural network converts non-binary data in all driving data into binary data by adopting an independent encoder;
s1.2: then, converting each item of driving data into a specified detection interval through normalization processing, dividing each item of driving data into a verification set, a test set and a training set, and repeatedly using each group of data in the verification set for multiple times to verify the driving stability of the patrol car;
s1.3: the method comprises the steps of counting root mean square errors of data in a test set, carrying out primary prediction on each group of data, outputting data with the best prediction result as optimal parameters, carrying out standardized processing on a training set according to the optimal parameters to generate training samples, then conveying the training samples to an optimized neural network, carrying out real-time optimization on the driving efficiency of the patrol car by adopting a long-term iteration method, and calculating a loss value of the patrol car in the driving process through a focus loss function;
s1.4: and the parameter analysis module receives the calculated loss values of each group, analyzes the accuracy, the detectable rate and the false alarm rate, feeds back the analysis result to a worker for analysis and adjustment, constructs an upper layer controller and a lower layer controller at the same time, generates drive control and brake control based on a fuzzy control theory and longitudinal dynamics, and optimizes the parameters of the patrol car according to the generated drive control and brake control.
As a further scheme of the present invention, the data recovery of the built-in storage module specifically comprises the following steps:
s2.1: the control platform periodically calculates the recovery rate of the temporarily stored data of the built-in storage module according to a system default or manually set cycle time value, and sends a recovery instruction to the built-in storage module;
s3.1: the built-in storage module is started after receiving the receiving instruction, receives the recovery rate numerical value calculated by the control platform, extracts each group of temporary storage data, recovers the temporary storage data according to the calculated recovery rate, generates a recovery log, records the recovery data information and the recovery date in the recovery log, and feeds the recovery data information and the recovery date back to the control platform for a manager to check.
Compared with the prior art, the invention has the beneficial effects that:
1. the patrol car is provided with a target detection module, a patrol car acquires surrounding environment image data through a vehicle-mounted camera, the target detection module constructs a primary target detection network to carry out frame-by-frame extraction of patrol pictures on the image data, then scales the acquired groups of environment patrol pictures according to different resolutions set by an administrator, extracts characteristic information of each group of patrol pictures in a picture data set, processes each group of characteristic information, outputs a detection frame, collects detection frame information and generates corresponding detection frame coordinates, carries out enlarged cutting on the related patrol pictures, collects each group of enlarged cut object pictures, then constructs a secondary target detection network to filter out simple negative samples belonging to a background in each group of object pictures, selects areas possibly containing objects to carry out classification and regression, then carries out detection on object positions in each group of object pictures through enlarged cutting, carries out cascade analysis on the acquired image data through constructing the primary target detection network and the secondary target detection network, can more accurately analyze each object information in the patrol car environment, improves the accuracy of patrol car target detection, and reduces the work load of related personnel;
2. the system is provided with a built-in storage module, a target detection module guides each group of enlarged and cut object pictures into the built-in storage module for temporary storage, a control platform periodically calculates the recovery rate of temporarily stored data of the built-in storage module according to a cycle time value which is default or manually set by a system, sends a recovery instruction to the built-in storage module, the built-in storage module is started after receiving the recovery instruction, receives the recovery rate value calculated by the control platform, extracts each group of temporarily stored data, recovers the temporarily stored data according to the calculated recovery rate, generates a recovery log, records the recovery data information and the recovery date in the recovery log, and feeds the recovery data information and the recovery date back to the control platform for a manager to check, so that the data stored in the system can be automatically recovered and deleted, the utilization rate of the system memory is improved, the redundant data is prevented from being excessively reduced, the communication transmission efficiency is prevented, and the system is ensured to operate stably.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.
Fig. 1 is a system block diagram of a remote driving patrol car system according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Example 1
Referring to fig. 1, a remote driving patrol car system comprises a control platform, a vehicle-mounted camera, a communication connection module, a target detection module, a parameter adjustment module, a fault detection module, a built-in storage module, a cloud database and an alarm module.
And the control platform is used for a manager to log in and control the patrol car to run, and meanwhile, feedback information of each submodule is received and is provided for the manager to check in a graphical mode.
The vehicle-mounted camera is used for acquiring surrounding environment information of the patrol car so as to generate image data.
And the communication connection module is used for receiving the control instruction sent by the control platform and carrying out communication detection.
Specifically, the communication connection module creates a group of sockets in the form of file descriptors, binds the generated sockets with the local IP address port to set monitoring, and after the monitoring is successfully set, the communication connection module waits for receiving a connection request of a client, acquires a group of new file descriptors, and calls an accept function to detect a read buffer of the file descriptors, and if the read buffer does not detect data, the function blocks and disconnects the connection, and if the read buffer detects data, the function blocks and establishes the connection.
The target detection module is used for receiving the image data and analyzing and feeding back the image data.
Specifically, the target detection module processes video or image sequence frames with a fixed frame rate in image data, then calculates and records the interval time of actual video frames, simultaneously constructs a primary target detection network to extract inspection pictures frame by frame from the image data, constructs a picture data set according to different sizes of each group of inspection pictures, then scales each group of acquired environment inspection pictures according to different resolutions set by an administrator, constructs an image pyramid according to the scaled inspection pictures, extracts characteristic information of each group of inspection pictures in the picture data set, sends each group of characteristic information into a bidirectional characteristic pyramid for characteristic fusion, performs scale normalization processing on the images by the image pyramid, and outputs detection frames, categories and scores, the target detection module collects object detection frame information in the inspection pictures and generates corresponding detection frame coordinates, the related routing inspection pictures are enlarged and cut, each group of enlarged and cut object pictures are collected and temporarily stored through a built-in storage module, then a secondary target detection network is constructed to filter simple negative samples belonging to the background in each group of object pictures, regions possibly containing objects are selected for classification and regression, then nine anchor frames with specified proportion and size are generated at each point of five object pictures with high semantic information and low semantic information and are classified and regressed, the positions of the objects in each group of enlarged object pictures are detected through the enlarged cutting, the detection results are fed back to a control platform, simultaneously target tracking is carried out, a primary target detection network and a secondary target detection network are constructed to carry out cascade analysis on the acquired image data, and each object information in the routing environment can be more accurately analyzed, the accuracy of the patrol car target detection is improved, the analysis efficiency of related personnel is improved, and the workload of the related personnel is reduced.
It is further noted that the target detection module establishes a motion model according to a kalman filtering theory, and simultaneously obtains a motion state of each object in real time through the established motion model, then the motion model defines the motion state of each object in a video frame according to a linear motion hypothesis of each object, collects the motion state of each object in the current video frame, establishes a prediction equation to estimate the motion state of each object in the next video frame, calculates a cosine distance between the detection result and the object according to detection results of all objects in the current video frame of the i-th vehicle-mounted camera calculated by a multi-target real-time detection algorithm, filters through a manually set threshold to obtain a cost matrix of cosine distance representation, then performs binary matching to obtain a matching result set, then the target detection module updates gain and covariance matrix in the motion model of the object that finally fails to be matched, re-matches each group of objects that finally fails to be matched, after matching is completed, sequentially processes video frames of each path of video stream in parallel, and sequentially performs cross-sampling and cross-tracking motion state estimation on the objects and monitoring object frame data obtained after the matching is completed.
Example 2
Referring to fig. 1, a remote driving patrol car system comprises a control platform, a vehicle-mounted camera, a communication connection module, a target detection module, a parameter adjustment module, a fault detection module, a built-in storage module, a cloud database and an alarm module.
The parameter adjusting module is used for collecting driving parameters of the patrol car and then carrying out optimization adjustment on the patrol car.
The method comprises the steps that a parameter adjusting module receives various driving data of a patrol car, an optimized neural network corresponding to the patrol car is constructed, the optimized neural network converts non-binary data in the driving data into binary data by adopting an independent encoder, the driving data are converted into a specified detection interval through normalization processing, the driving data are divided into a verification set, a test set and a training set, meanwhile, various groups of data in the verification set are repeatedly used for verifying the driving stability of the patrol car, root mean square errors of the data in the test set are counted, each group of data are predicted once, the data with the best prediction result are output as optimal parameters, the training set is subjected to standardization processing according to the optimal parameters to generate training samples, the training samples are conveyed into the optimized neural network, the driving efficiency of the patrol car is optimized in real time by adopting a long-term iteration method, the loss value of the driving process of the patrol car is calculated through a focus loss function, finally, the parameter analyzing module receives the calculated loss values of the groups, analyzes the loss values and the analysis rate of the working efficiency of the patrol car, and feeds back the analysis results to a patrol car working controller for constructing, and controlling the patrol car according to the theoretical driving parameters, and the generation of a driving control of a driving mechanism.
The fault detection module is used for receiving the driving data of the patrol car and feeding back the fault.
The built-in storage module is used for temporarily storing each group of data of the patrol car and periodically recovering the data.
Specifically, the control platform periodically calculates the recovery rate of the temporary storage data of the built-in storage module according to a cycle time value which is set by default or manually, and sends a recovery instruction to the built-in storage module, the built-in storage module is started after recovering the recovery instruction, and receives the recovery rate value calculated by the control platform, then extracts each group of temporary storage data, and recovers the temporary storage data according to the calculated recovery rate, and then generates a recovery log, and records the recovery data information and the recovery date in the recovery log, and feeds the recovery data information and the recovery date back to the control platform for a manager to check, so that the data stored in the control platform can be automatically recovered and deleted, the memory utilization rate of the system is improved, the redundant data is prevented from reducing the communication transmission efficiency too much, and the operation stability of the system is ensured.
The cloud database is used for receiving and storing each group of data uploaded by the control platform.
And the alarm module is used for receiving each group of data fed back by the target detection module and sending alarm information to a manager according to the data content.
Claims (6)
1. A remote driving patrol car system is characterized by comprising a control platform, a vehicle-mounted camera, a communication connection module, a target detection module, a parameter adjustment module, a fault detection module, a built-in storage module, a cloud database and an alarm module;
the control platform is used for a manager to log in and control the patrol car to run, and meanwhile, feedback information of each sub-module is received and is provided for the manager to check in a graphical mode;
the vehicle-mounted camera is used for acquiring surrounding environment information of the patrol car to generate image data;
the communication connection module is used for receiving a control instruction sent by the control platform and carrying out communication detection;
the target detection module is used for receiving the image data and analyzing and feeding back the image data;
the parameter adjusting module is used for collecting the driving parameters of the patrol car and then carrying out optimization adjustment on the patrol car;
the fault detection module is used for receiving the driving data of the patrol car at any time and carrying out fault feedback;
the built-in storage module is used for temporarily storing each group of data of the patrol car and periodically recovering the data;
the cloud database is used for receiving and storing each group of data uploaded by the control platform;
and the alarm module is used for receiving each group of data fed back by the target detection module and sending alarm information to a manager according to the data content.
2. The remote driving patrol car system according to claim 1, wherein the communication connection module specifically comprises the following steps:
the method comprises the following steps: the communication connection module creates a group of sockets in the form of file descriptors and binds the generated sockets with a local IP address port to set monitoring;
step two: after the monitoring setting is successful, the communication connection module waits for and receives a connection request of a client, acquires a group of new file descriptors and calls an accept function to detect a read buffer area of the file descriptors;
step three: if the read buffer area can not detect the data, the function is blocked and the connection is disconnected, and if the read buffer area detects the data, the function is unblocked and the connection is established.
3. The remote driving patrol car system according to claim 1, wherein the target detection module analyzes feedback by the following specific steps:
step (1): the target detection module processes video or image sequence frames with a fixed frame rate in the image data, then calculates and records the interval time of actual video frames, simultaneously constructs a primary target detection network to extract inspection pictures from the image data frame by frame, and constructs a picture data set according to different sizes of each group of inspection pictures;
step (2): then, zooming each group of acquired environment inspection pictures according to different resolutions set by an administrator, constructing an image pyramid according to the zoomed inspection pictures, extracting feature information of each group of inspection pictures in the picture data set, sending each group of feature information into a bidirectional feature pyramid for feature fusion, carrying out scale normalization processing on the feature information by the image pyramid, and outputting a detection frame, a category and a score;
and (3): collecting object detection frame information in the inspection pictures, generating corresponding detection frame coordinates, performing expanded cutting on related inspection pictures, collecting each group of enlarged and cut object pictures, performing temporary storage through a built-in storage module, constructing a secondary target detection network, filtering out simple negative samples belonging to a background in each group of object pictures, and selecting areas possibly containing objects for classification and regression;
and (4): nine anchor frames with specified proportion and size are generated at each point of the five object pictures with high or low semantic information, the anchor frames are classified and regressed, the object positions in each group of object pictures are detected through enlarged cutting, the detection results are fed back to the control platform, and meanwhile target tracking is carried out.
4. A remote driving patrol car system according to claim 3, wherein the target tracking in the step (4) comprises the following specific steps:
the first step is as follows: establishing a motion model through a Kalman filtering theory, simultaneously acquiring the motion state of each object in real time through the established motion model, and then defining the motion state of each object in a video frame by the motion model according to a linear motion hypothesis of each object;
the second step: collecting the motion state of each object in the current video frame, constructing a prediction equation to estimate the motion state of each object in the next video frame, and calculating the detection results of all objects in the current video frame of the ith vehicle-mounted camera according to a multi-target real-time detection algorithm;
the third step: calculating cosine distance between the detection result and the object, filtering through a manually set threshold value to obtain a cost matrix represented by the cosine distance, and then performing binary matching to obtain a matching result set;
the fourth step: and the target detection module updates Kalman gain and covariance matrixes in the motion model of the object which finally fails to be matched, matches each group of objects which finally fails to be matched again after the updating is finished, processes video frame data of each path of video stream in parallel after the matching is finished, and sequentially executes object marking, estimation of the motion state of the object, matching correlation and cross-monitoring multi-object tracking on video frames obtained after down-sampling in each path of video stream.
5. The remote driving patrol car system according to claim 1, wherein the parameter adjusting module comprises the following specific steps:
s1.1: the parameter adjusting module receives all driving data of the patrol car and constructs an optimized neural network corresponding to the patrol car, and the optimized neural network adopts an independent encoder to convert non-binary data in all driving data into binary data;
s1.2: converting all the driving data into a specified detection interval through normalization processing, dividing all the driving data into a verification set, a test set and a training set, and repeatedly using all the groups of data in the verification set for multiple times to verify the driving stability of the patrol car;
s1.3: the method comprises the steps of counting root mean square errors of data in a test set, carrying out primary prediction on each group of data, outputting data with the best prediction result as optimal parameters, carrying out standardized processing on a training set according to the optimal parameters to generate training samples, then conveying the training samples to an optimized neural network, carrying out real-time optimization on the driving efficiency of the patrol car by adopting a long-term iteration method, and calculating a loss value of the patrol car in the driving process through a focus loss function;
s1.4: and the parameter analysis module receives the calculated loss values of each group, analyzes the accuracy, the detectable rate and the false alarm rate, feeds back the analysis result to a worker for analysis and adjustment, constructs an upper layer controller and a lower layer controller at the same time, generates drive control and brake control based on a fuzzy control theory and longitudinal dynamics, and optimizes the parameters of the patrol car according to the generated drive control and brake control.
6. The remote driving patrol car system according to claim 1, wherein the data recovery of the built-in storage module comprises the following specific steps:
s2.1: the control platform periodically calculates the recovery rate of the temporarily stored data of the built-in storage module according to a system default or manually set cycle time value, and sends a recovery instruction to the built-in storage module;
s3.1: the built-in storage module is started after receiving the receiving instruction, receives the recovery rate numerical value calculated by the control platform, extracts each group of temporary storage data, recovers the temporary storage data according to the calculated recovery rate, generates a recovery log, records the recovery data information and the recovery date in the recovery log, and feeds the recovery data information and the recovery date back to the control platform for a manager to check.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210967369.4A CN115409992A (en) | 2022-08-12 | 2022-08-12 | Remote driving patrol car system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210967369.4A CN115409992A (en) | 2022-08-12 | 2022-08-12 | Remote driving patrol car system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115409992A true CN115409992A (en) | 2022-11-29 |
Family
ID=84158922
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210967369.4A Pending CN115409992A (en) | 2022-08-12 | 2022-08-12 | Remote driving patrol car system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115409992A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116189439A (en) * | 2023-05-05 | 2023-05-30 | 成都市青羊大数据有限责任公司 | Urban intelligent management system |
| TWI823819B (en) * | 2023-05-15 | 2023-11-21 | 先進車系統股份有限公司 | Driving assistance system and driving assistance computation method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215169A (en) * | 2017-07-04 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | Storage method, device and the equipment of travelling data |
| CN109541943A (en) * | 2018-12-07 | 2019-03-29 | 西南交通大学 | A kind of tramcar on-line optimizing and controlling method |
| CN110446180A (en) * | 2019-07-02 | 2019-11-12 | 杭州律橙电子科技有限公司 | A kind of vehicle intelligent terminal system and its operating method |
| CN111476826A (en) * | 2020-04-10 | 2020-07-31 | 电子科技大学 | Multi-target vehicle tracking method based on SSD target detection |
| CN114394100A (en) * | 2022-01-12 | 2022-04-26 | 深圳力维智联技术有限公司 | Unmanned prowl car control system and unmanned car |
| CN114779794A (en) * | 2022-06-21 | 2022-07-22 | 东风悦享科技有限公司 | Street obstacle identification method based on unmanned patrol vehicle system in typhoon scene |
-
2022
- 2022-08-12 CN CN202210967369.4A patent/CN115409992A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215169A (en) * | 2017-07-04 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | Storage method, device and the equipment of travelling data |
| CN109541943A (en) * | 2018-12-07 | 2019-03-29 | 西南交通大学 | A kind of tramcar on-line optimizing and controlling method |
| CN110446180A (en) * | 2019-07-02 | 2019-11-12 | 杭州律橙电子科技有限公司 | A kind of vehicle intelligent terminal system and its operating method |
| CN111476826A (en) * | 2020-04-10 | 2020-07-31 | 电子科技大学 | Multi-target vehicle tracking method based on SSD target detection |
| CN114394100A (en) * | 2022-01-12 | 2022-04-26 | 深圳力维智联技术有限公司 | Unmanned prowl car control system and unmanned car |
| CN114779794A (en) * | 2022-06-21 | 2022-07-22 | 东风悦享科技有限公司 | Street obstacle identification method based on unmanned patrol vehicle system in typhoon scene |
Non-Patent Citations (1)
| Title |
|---|
| 顾超越等: ""基于改进Faster-RCNN的无人机巡检架空线路销钉缺陷检测"", 《高电压技术》, 30 September 2020 (2020-09-30), pages 1 - 2 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116189439A (en) * | 2023-05-05 | 2023-05-30 | 成都市青羊大数据有限责任公司 | Urban intelligent management system |
| TWI823819B (en) * | 2023-05-15 | 2023-11-21 | 先進車系統股份有限公司 | Driving assistance system and driving assistance computation method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353413B (en) | Low-missing-report-rate defect identification method for power transmission equipment | |
| CN110084165B (en) | Intelligent identification and early warning method for abnormal events in open scene of power field based on edge calculation | |
| CN110826538B (en) | Abnormal off-duty identification system for electric power business hall | |
| CN115409992A (en) | Remote driving patrol car system | |
| CN111080693A (en) | Robot autonomous classification grabbing method based on YOLOv3 | |
| CN103020590B (en) | A kind of vehicle identification system based on three-dimensional model and images match and method thereof | |
| CN110532937B (en) | Method for accurately identifying forward targets of train based on identification model and classification model | |
| CN110648319B (en) | Equipment image acquisition and diagnosis system and method based on double cameras | |
| CN113284144B (en) | Tunnel detection method and device based on unmanned aerial vehicle | |
| CN113312957A (en) | off-Shift identification method, device, equipment and storage medium based on video image | |
| CN112349057A (en) | Deep learning-based indoor smoke and fire detection method | |
| CN111881958A (en) | License plate classification recognition method, device, equipment and storage medium | |
| CN112836683A (en) | License plate recognition method, device, equipment and medium for portable camera equipment | |
| CN113408630A (en) | Transformer substation indicator lamp state identification method | |
| CN210666820U (en) | Pedestrian abnormal behavior detection system based on DSP edge calculation | |
| CN115457411B (en) | Unmanned inspection method and device for aviation oil pipeline and aviation oil safety operation and maintenance system | |
| CN108596068B (en) | Method and device for recognizing actions | |
| CN111413691A (en) | Semantic positioning and mapping method adopting distributed structure | |
| CN116229424A (en) | Intelligent networking automobile pedestrian detection method based on improved YOLO v3 | |
| CN114529816A (en) | Tea garden pest and disease monitoring system based on image recognition | |
| CN112819988A (en) | Unmanned aerial vehicle power station intelligent inspection method and system based on 5G and network side server | |
| CN113610191B (en) | Garbage classification model modeling method and garbage classification method | |
| CN117610971B (en) | Highway electromechanical system health index evaluation method | |
| CN113887718B (en) | Channel pruning method and device based on relative activation rate and lightweight flow characteristic extraction network model simplification method | |
| US20230109494A1 (en) | Methods and devices for building a training dataset |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |