CN104442631A - Intelligent automotive user manual interaction method - Google Patents

Abstract

The invention belongs to the field of automotive technologies, and particularly relates to an intelligent automotive user manual interaction method. The control process includes the steps that step 1, images for indicating icons on an automobile are collected; step 2, the collected images are detected to obtain interesting regions; step 3, PCA-HOG feature extraction is conducted on all the interesting regions, online recognition is conducted on extracted features through a pre-trained support vector machine classifier, and the exact meaning of the collected images is obtained; step 4, function demonstration or broadcasting is conducted on the information related to the collected icon images and stored in an automotive body controller. According to the intelligent automotive user manual interaction method, a mobile phone is adopted to collection the information, the demands of users are recognized through interaction of an image recognition technology and a processing technology, the operation of corresponding functions or icon meaning is demonstrated to the users according to the demands of the users, intelligentizing of a user manual is achieved, and convenience in consulting the user manual by the users is improved greatly.

Description

A kind of automobile intelligent user manual exchange method

Technical field

The invention belongs to automobile technical field, be specifically related to a kind of automobile intelligent user manual exchange method.

Background technology

Along with the development of automotive technology, the function of automobile becomes increasingly abundant, automobile operator's hand book is the important means that user understands automobile function, existing automobile operator's hand book has papery version and electronic edition, papery version is that user manually browses handbook inquiry, and electronic edition is that user manually inputs the information that need understand and inquires about.But increasing along with automobile function, automobile operator's hand book also can constantly upgrade, operator's hand book is just caused to become increasingly complex, increase the difficulty that user understands automobile, for user, no matter be papery version, or the operator's hand book of electronic edition, wanting very quick, the convenient end, to consult certain function convenient all not.

Summary of the invention

Object of the present invention is exactly the deficiency existed to solve above-mentioned background technology, provide one simply, automobile intelligent user manual exchange method efficiently.

The technical solution used in the present invention is: a kind of automobile intelligent user manual exchange method, comprises the following steps:

Step 1, collection vehicle indicates the image of icon, outputs to car body controller;

Step 2, the image of car body controller to input detects, obtain area-of-interest: the image of car body controller to input carries out binary conversion treatment, edge extracting is carried out to the binary image after process, obtain the edge contour data of indicator chart picture, determine the area-of-interest at icon place according to the shape of edge contour data and textural characteristics;

Step 3, carries out PCA-HOG feature extraction respectively to all area-of-interests, and the feature training in advance of extraction support vector machine classifier is out carried out ONLINE RECOGNITION, obtains the accurate meaning gathering image;

Step 4, carries out function presentation or report by the information relevant to the icon image gathered stored in car body controller.

Further, by mobile phone camera collection vehicle indicating the image of icon in described step 1, the image of collection is transferred to car body controller by bluetooth or WIFI by mobile phone.

Further, describedly to the process that area-of-interest carries out PCA-HOG feature extraction be:

1) area-of-interest of acquisition is normalized to the image I ∈ R of m pixel × n-pixel ^{m × n}, I (x, y) represents the arbitrary pixel in image I, then can calculate gradient magnitude and the gradient direction of image I,

2) image I is evenly divided into s _w× s _hindividual mesh region, is defined as a Cell by each region, the gradient direction of pixel in each Cell is divided in the scope of 0-180 ° 9 bin (i.e. every 20 ° of bin), and calculates the gradient magnitude histogram Hcell of each bin;

3) adjacent every 2 × 2 Cell are combined as a Block, adopt partly overlapping mode to arrange between adjacent Block, overlapping region is 2 Cell;

4) histogram of each Cell of each Block will be formed according to from left to right, combine from top to bottom, form the histogram Hblock of a Block, and in Block, using L2 norm normalization histogram to eliminate the impact of illumination, the histogram of the Block after normalization method represents with NHblock;

5) calculate nonlinear principal component component, make Γ ∈ R ^{np × nf}, represent that the HOG in each Block region describes the front n of operator _pindividual major component, describes operator H and projects to the linear subspaces be made up of major component Γ by HOG;

6) whole detection window is divided into n Block, the set of histograms of all Block is just constituted altogether the HOG feature of this detection window.

Further, the training process of described support vector machine classifier is: carry out feature extraction respectively to each icon sample in advance, set the unique features of each sample.

The present invention utilizes smart mobile phone Information Monitoring, undertaken alternately by image recognition technology and treatment technology and user, identify the demand of user, and demonstrate the operation of corresponding function or the implication of icon according to the demand of user to user, realize the intellectuality of user manual, the user of great raising consults the convenience of user manual, solves user well and consults papery user manual or electronic documentation interaction is poor, uses problem not easily.

Accompanying drawing explanation

Fig. 1 is the present invention based on hardware platform system schematic diagram.

Fig. 2 is control flow schematic diagram of the present invention.

Fig. 3 is the schematic diagram of icon detection and Identification in Fig. 2.

Detailed description of the invention

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail, is convenient to be well understood to the present invention, but they do not form restriction to the present invention.

As shown in Figure 1, intelligent subscriber handbook exchange method of the present invention realizes with software system based on existing hardware platform system.Wherein hardware platform system is made up of car body controller 1, voice system 3, read-out 4, loud speaker 5, smart mobile phone 6 etc., software system are apply the related software that mobile phone and car body controller can be made to contact, and intelligent handbook 2 is for being stored in the icon data in car body controller.Car body controller 1 is connected with loud speaker 5 by wire harness, and mobile phone 6 carries out data interaction by USB, WIFI, bluetooth and main frame.Car body controller hardware components is mainly divided into: data processing module, interface module, communication module, memory module.The graphicinformation acting as calculating touch-screen feedack and gathered by mobile phone of data processing unit; Memory cell act as the electronic document, operation demonstration video and the information of voice prompt that store user vehicle handbook; Interface module comprises the interface of touch LCD screen, the interface of loud speaker; Communication module mainly comprises USB, WIFI and bluetooth.

The present invention utilizes image recognition technology to realize the mutual of user and intelligent handbook 2, by the function key on the camera scanning car of smart mobile phone 6, (instrument on instruction icon on knob or each read-out, air-conditioning panel, middle control screen and the audio-visual amusement read-out of rear row etc.) instruction icon, by being contained in the application program of mobile phone, the image of the function icon collected by mobile phone camera is transferred to car body controller by bluetooth or WIFI, car body controller detects the icon scanned, identify, finally according to the result identified, the mode of voice broadcast is adopted to instruct being identified the use of function or pointing out the implication of this icon, the mode of image or video playback can be adopted on mobile phone or on car simultaneously, point out the using method of this function or the implication of icon.

As described in Fig. 2, Fig. 3, the concrete interaction flow of automobile intelligent user manual is:

Scanning icon: by the image of the instruction icon of (instrument, air-conditioning panel, middle control screen and the audio-visual amusement read-out of rear row etc.) on the function key in the camera collection automobile of smart mobile phone, instruction icon on knob or each read-out, by the WIFI of mobile phone or Bluetooth transmission to the car body controller on automobile.

Icon detects: after car body controller receives the image of smart mobile phone transmission, the area-of-interest that doubtful icon exists is obtained according to the distinctive feature of instruction icon (shape, color, texture etc.) in the function icon of automobile or read-out, be specially: binary conversion treatment is carried out to the image of input, edge extracting is carried out to the binary image after process, obtain the edge contour data of indicator chart picture, determine the area-of-interest at icon place according to the shape of edge contour data and textural characteristics.

Image recognition: feature extraction is carried out to the above-mentioned area-of-interest detected, then the good segregator of training in advance is used to carry out Classification and Identification to these features, thus obtain the definite implication of above-mentioned scanning icon: respectively PCA-HOG feature extraction is carried out to all area-of-interests, the feature training in advance of extraction support vector machine classifier is out carried out ONLINE RECOGNITION, and the accurate meaning obtaining gathering image exports to function report module.

The process of area-of-interest being carried out to PCA-HOG feature extraction is:

1) area-of-interest of acquisition is normalized to the image I ∈ R of m pixel × n-pixel ^{m × n}, I (x, y) represents the arbitrary pixel in image I, then can calculate gradient magnitude and the gradient direction of image I,

The gradient magnitude of I (x, y) is: $g(x,y)=\sqrt{I_x{(x,y)}^2+I_y{(x,y)}^2}$

The gradient direction of I (x, y) is: $\mathrm{\θ}(x,y)=\arctan \frac{I_y(x,y)}{I_x(x,y)};$

2) image I is evenly divided into s _w× s _hindividual mesh region, each region is defined as a Cell, the gradient direction of pixel in each Cell is divided in the scope of 0-180 ° 9 bin (i.e. every 20 ° of bin), and calculates the gradient magnitude histogram Hcell of each bin, wherein i represents i-th unit

Hcell _i(k)＝Hcell _i(k)+g(x,y) I(x,y)∈cell _i&&θ(x,y)∈bin(k)；

3) adjacent every 2 × 2 Cell are combined as a Block, adopt partly overlapping mode to arrange between adjacent Block, overlapping region is 2 Cell;

4) histogram of each Cell of each Block will be formed according to from left to right, combine from top to bottom, form the histogram Hblock of a Block, and in Block, use L2 norm normalization histogram to eliminate the impact of illumination, the histogram of the Block after normalization method represents with NHblock, Hblock _j={ Hcell ₁, Hcell ₂, Hcell ₃, Hcell ₄}

5) calculate nonlinear principal component component, make Γ ∈ R ^{np × nf}, represent that the HOG in each Block region describes the front n of operator _pindividual major component, describes operator H and projects to the linear subspaces be made up of major component Γ by HOG;

6) whole detection window is divided into n Block, the set of histograms of all Block is just constituted altogether the HOG feature of this detection window, f _hog={ Yblock ₁, Yblock ₂..., Yblock _n.

The training process of above-mentioned support vector machine classifier, for carry out feature extraction respectively to each icon sample in advance, namely sets learning process in advance, obtains the specific feature of each icon sample.The process of feature extraction is herein substantially identical with above-mentioned PCA-HOG characteristic extraction procedure, difference is it is directly carry out feature extraction to the known icon sample gathered herein, use SVM classifier to train the feature of extraction, the result of training is the segregator of Based PC A-HOG feature; And above-mentioned be that feature extraction is carried out to the area-of-interest of the unknown, after feature extraction is carried out to area-of-interest, the segregator of the characteristic area of extraction by Based PC A-HOG feature is identified, draw scanning icon accurate meaning.

Function presentation/report: according to the result of icon identification, utilizes the read-out on automobile, helps the use of user's decorrelation function or the implication of icon by the mode of displaying video, image and word; Help the use of user's decorrelation function or the implication of icon by the mode of the speaker sound play cuing in car simultaneously; And the mode of image, video playback or voice broadcast can be adopted on mobile phone, point out the using method of this function or the implication of icon.

The content be not described in detail in this specification sheets belongs to the known prior art of professional and technical personnel in the field.

Claims (4)

1. an automobile intelligent user manual exchange method, is characterized in that comprising the following steps:

Step 1, collection vehicle indicates the image of icon, outputs to car body controller;

Step 2, the image of car body controller to input detects, obtain area-of-interest: the image of car body controller to input carries out binary conversion treatment, edge extracting is carried out to the binary image after process, obtain the edge contour data of indicator chart picture, determine the area-of-interest at icon place according to the shape of edge contour data and textural characteristics;

Step 3, carries out PCA-HOG feature extraction respectively to all area-of-interests, and the feature training in advance of extraction support vector machine classifier is out carried out ONLINE RECOGNITION, obtains the accurate meaning gathering image;

Step 4, carries out function presentation or report by the information relevant to the icon image gathered stored in car body controller.

2. a kind of automobile intelligent user manual exchange method according to claim 1, it is characterized in that: by mobile phone camera collection vehicle indicating the image of icon in described step 1, the image of collection is transferred to car body controller by bluetooth or WIFI by mobile phone.

3. a kind of automobile intelligent user manual exchange method according to claim 1, is characterized in that: describedly to the process that area-of-interest carries out PCA-HOG feature extraction be:

1) area-of-interest of acquisition is normalized to the image I ∈ R of m pixel × n-pixel ^{m × n}, I (x, y) represents the arbitrary pixel in image I, then can calculate gradient magnitude and the gradient direction of image I,

2) image I is evenly divided into s _w× s _hindividual mesh region, is defined as a Cell by each region, the gradient direction of pixel in each Cell is divided in the scope of 0-180 ° 9 bin, and calculates the gradient magnitude histogram Hcell of each bin;

3) adjacent every 2 × 2 Cell are combined as a Block, adopt partly overlapping mode to arrange between adjacent Block, overlapping region is 2 Cell;

4) histogram of each Cell of each Block will be formed according to from left to right, combine from top to bottom, form the histogram Hblock of a Block, and in Block, using L2 norm normalization histogram to eliminate the impact of illumination, the histogram of the Block after normalization method represents with NHblock;

5) calculate nonlinear principal component component, make Γ ∈ R ^{np × nf}, represent that the HOG in each Block region describes the front n of operator _pindividual major component, describes operator H and projects to the linear subspaces be made up of major component Γ by HOG;

6) whole detection window is divided into n Block, the set of histograms of all Block is just constituted altogether the HOG feature of this detection window.

4. a kind of automobile intelligent user manual exchange method according to claim 1, is characterized in that: the training process of described support vector machine classifier is: carry out feature extraction respectively to each icon sample in advance, set the unique features of each sample.