CN112351210A

CN112351210A - Active vision acquisition system

Info

Publication number: CN112351210A
Application number: CN202011228762.9A
Authority: CN
Inventors: 朱海龙; 贺维; 张广玲
Original assignee: Harbin Normal University
Current assignee: Harbin Normal University
Priority date: 2020-11-06
Filing date: 2020-11-06
Publication date: 2021-02-09

Abstract

The invention discloses an active vision acquisition system which comprises 1 active vision control module and n video acquisition sensors, wherein an active vision control algorithm is embedded in the active vision control module, and a control instruction is sent to the n video acquisition sensors in a wireless mode, so that the video acquisition module has activity when acquiring video data, and the direction angle, the pitch angle and the focal length can be automatically adjusted as required. The traditional passive video acquisition mode is converted into an anthropomorphic active visual model, namely: after an interested target is found, the active vision control module dispatches the full-field video sensor to acquire video data with better quality from different angles and focal lengths for analysis and use.

Description

Active vision acquisition system

Technical Field

The invention relates to an acquisition system, in particular to an active vision acquisition system.

Background

Video capture handle simulationVideo conversionBecome intoDigital videoAnd saved in the format of the digital video file. The video acquisition is toAnalog cameraThe video signal output by video recorder, LD video disk player and TV set is converted into binary digital information by special analog and digital conversion equipment.

The traditional passive video acquisition mode is relatively slow in acquisition of a target and cannot acquire high-quality videos from multiple angles.

Disclosure of Invention

The invention aims to provide an active visual acquisition system, which is converted from a traditional passive video acquisition mode into an anthropomorphic active visual model, namely: after an interested target is found, the active vision control module dispatches the full-field video sensor to acquire video data with better quality from different angles and focal lengths for analysis and use, so that the problems in the background technology can be solved.

In order to achieve the purpose, the invention provides the following technical scheme:

the utility model provides an initiative vision acquisition system, includes 1 initiative vision control module and n video acquisition sensors, the initiative vision control module is embedded to have initiative vision control algorithm, sends control command to n video acquisition sensors through wireless mode for video acquisition module has the initiative when gathering video data, can automatic adjustment direction angle, every single move angle and focus as required.

Furthermore, the active vision control module can control n video acquisition sensors to work independently according to needs, or multiple machines cooperate to acquire video data of the same interested target from different angles and heights so as to acquire the optimal target characteristic data.

Furthermore, the active visual control module can be networked with the data processing center, can upload data as a component of the global planning of the data processing center, and can also receive a control strategy downloaded by the data processing center and update the scheduling method of the video sensor of the current scene.

Furthermore, each video acquisition sensor is provided with a 3D cloud platform and has an adjustable focal length, and a wired or wireless mode can be adopted according to actual conditions when video data are uploaded.

Further, the active visual control algorithm comprises a structural traffic scene modeling, a moving target track tracking model in a structural scene and a visual sensor scheduling model.

Compared with the prior art, the invention has the beneficial effects that:

the traditional passive video acquisition mode is converted into an anthropomorphic active visual model, namely: after an interested target is found, the active vision control module dispatches the full-field video sensor to acquire video data with better quality from different angles and focal lengths for analysis and use.

Drawings

FIG. 1 is a block diagram of the system components of the present invention;

FIG. 2 is a schematic diagram of a rotary and a structured scene model according to the present invention;

FIG. 3 is a diagram of the trajectory of a motor vehicle entering a rotary island according to the present invention;

FIG. 4 is a diagram illustrating simulation experiment results of shooting angles according to the present invention;

FIG. 5 is a structured scene model of three sensors of the present invention;

FIG. 6 is a graph showing the experimental results of FIG. 5 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. 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 invention.

The utility model provides an initiative vision acquisition system, includes 1 initiative vision control module and n video acquisition sensors, the initiative vision control module is embedded to have initiative vision control algorithm, sends control command to n video acquisition sensors through wireless mode for video acquisition module has the initiative when gathering video data, can be as required automatic adjustment direction angle, every single move angle and focus, see figure 1 and show.

The active vision control module can control n video acquisition sensors to work independently according to needs, or multiple machines cooperate to acquire video data of the same interested target from different angles and heights so as to acquire the optimal target characteristic data.

The active visual control module can be networked with the data processing center, can upload data to be used as a component of global planning of the data processing center, and can also receive control strategies downloaded by the data processing center and update the scheduling method of the video sensor of the current scene.

Each video acquisition sensor has the 3D cloud platform and the focus is adjustable, can adopt wired or wireless mode according to actual conditions when uploading video data.

The active visual control algorithm comprises a structural traffic scene modeling, a moving target track tracking model in a structural scene and a visual sensor scheduling model.

1. Structured traffic scene modeling

The traffic parameters of the target scene are highly abstracted, a structured traffic scene model is constructed by combining with the actual engineering, and a highly digital scene model is provided for the subsequent video acquisition sensor.

Taking the rotary island as an example (see fig. 2), a structured scene model for active vision is established. In the process of establishing the roundabout model, contents such as the width of a motor vehicle lane, the speed of the motor vehicle, traffic rules for entering and exiting the roundabout and the like in the national road traffic standard are used as parameters of the model.

The core area of the rotary is assumed to be

Having a coordinate range of

(x-p_r)²+(y-q_r)²≤r_i ²(ii) a Road area coordinate range

r_i ²≤(x-p_r)²+(y-q_r)²≤r_o ². In the formula r_oRepresents the outer circle radius of the roundabout, r_iIs the radius of the circle in the rotary island, p_r，q_rRespectively the distance between the center of the circle and the x and y axes of the coordinate axis.

2. Active tracking model for moving target track in structured scene

The purpose of establishing the set of expected coordinates of the moving object is to provide pre-set coordinate parameters for the active visual scheduling algorithm.

When the motor vehicle enters the roundabout, the active visual model determines the tracking priority according to the emergency degree of the event, and actively tracks the track of the interested target. According to traffic regulations, vehicles enter the rotary island as shown in fig. 3.

Then the moving target moving just driving into the roundabout track model in the structured scene is shown as formula (1):

in the above formula (alpha)_e,β_e) Is the center of the ring island, p_e、q_eThe projection of the trajectory equation on the x and y axes and the initial point coordinate of the target vehicle entering the rotary island are shown. After entering the roundabout, the motion trajectory of the target vehicle can be predicted according to the model of fig. 2 as shown in equation (2):

(x-p_x)²+(y-p_y)²＝r_t,r_i+d_r＜x＜r_o-d_r，d_r＜y＜2r_o-d_r (2)

wherein d is_rVehicle width of moving object (p)_x,p_y) Is the center of the track circle. The expected coordinate set CR of the whole track of the interested moving vehicle target in the roundabout can be obtained by the formulas (1) and (2), and a data basis is provided for an active visual scheduling model of the next step.

The following is the video capture angle simulation experiment results, as shown in fig. 4.

TABLE 1 shooting angle variation trend chart

3. Video acquisition sensor scheduling model based on confidence rule base

Describing each constraint condition and parameter applicable condition in the structured scene as formula (3) by using a confidence rule, wherein a plurality of rules form a rule base so as to establish a scheduling model of the video acquisition sensor:

optimizing an active visual scheduling model in a structured scene on an acquired training sample set, wherein a mean square error is used as an objective function in the optimization process:

and a complete scheduling model is obtained after training, and when an event (such as violation, traffic accident and the like) occurs, a plurality of video acquisition sensors are scheduled to actively acquire data with higher quality and capture an accident scene, so that the duty assignment efficiency is improved, and the rescue time is shortened.

The experimental results are as follows.

Taking 3 video acquisition sensors as an example, 480 moving objects (including the situations of entering, exiting, winding around the island and the like at four intersection entering the island) are respectively scheduled, and are compared and analyzed with the accuracy index determined by the scheduling data with the label. The collection field and the complete structural scene are shown in fig. 5, and the positions of the video collection sensors are southwest corner, southeast corner and northeast corner respectively.

Experiments were performed using the K-means method (K-means) \ Fuzzy System (Fuzzy System), artificial neural Networks (neural Networks), Radial Basis functions (Radial Basis functions) and the method (Belief Rule Base), respectively, and labeled scheduling data (True Set) was used as a reference for comparison.

The results of the experiment are shown in FIG. 6. The experimental result is generated by comparing the calculation result of the active visual scheduling algorithm with the labeled scheduling data (True Set) as a standard. From experimental results (each method fitting curve is compared with a labeled scheduling data curve), the method is superior to other methods in stability, and is equivalent to other methods in precision.

The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be able to cover the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.

Claims

1. The utility model provides an initiative vision acquisition system, includes 1 initiative vision control module and n video acquisition sensor, its characterized in that, the initiative vision control module is embedded to have initiative vision control algorithm, sends control command to n video acquisition sensor through wireless mode for video acquisition module has the initiative when gathering video data, can be as required automatic adjustment direction angle, pitch angle and focus.

2. The active vision acquisition system of claim 1 wherein the active vision control module can control the n video acquisition sensors to work independently or multiple machines cooperate to acquire video data from different angles and heights for the same object of interest in order to obtain the best object feature data.

3. The active visual acquisition system of claim 1 wherein the active visual control module is networked with the data processing center, is capable of uploading data as part of a global planning for the data processing center, and is also capable of receiving control strategies downloaded by the data processing center to update the scheduling method for the video sensor of the current scene.

4. The active vision acquisition system of claim 1 in which each video acquisition sensor has a 3D pan/tilt and is adjustable in focal length, and uploading video data can be done in a wired or wireless manner depending on the actual situation.

5. The active visual acquisition system of claim 1 in which the active visual control algorithm comprises a structured traffic scene modeling, a moving object trajectory tracking model in a structured scene, a visual sensor scheduling model.