CN101867828B - Multi-charge coupled device group self-adaptive imaging method of cambered optical structure - Google Patents

Abstract

The invention relates to a multi-charge coupled device group self-adaptive imaging method of a cambered optical structure, belonging to a field of the image information acquisition and processing. By imitating an ommateum to adjust the nerve conduction mode with the change of the illumination under the condition of day alternates with night to obtain a self-adaptive imaging effect principle, the instrument uses a plurality of common charge coupled devices (CCD), corresponding auxiliary circuits and a neural processing algorithm to construct a biomimic self-adaptive adjustment imaging instrument. The imaging instrument has an imaging function of self adjusting fusion modes with the change of illumination conditions, can acquire a high-contrast image under a normal illumination condition, and also can acquire a high-brightness sensitive image under a weak illumination condition. The invention realizes high-definition surveillance and monitoring under the condition of day alternates with night, belonging to the image information acquisition and processing field.

Description

The multi-charge coupled apparatus group self adaptive imaging method of curved surface optical texture

One, technical field

The multi-charge coupled apparatus group self-adaptive imaging instrument and the method for curved surface optical texture of the present invention belong to image information acquisition and process field.

Two, background technology

In video scene supervision, commercial production monitoring etc. are used, not only need imageing sensor to have higher spatial resolution usually, also need imageing sensor under the darker situation of night or light, to get access to high-quality image information.Common imageing sensor all depends on extraneous illumination condition, and the reduction of lamp brightness certainly will have influence on imaging effect under night or other special weather situations.Though adopt at night usually and add the secondary light source of artificial light sources as imageing sensor; But the brightness range of this light source is limited; And the riding position of light source also can directly have influence on the imaging effect of transducer, causes characteristics such as image irradiation is inhomogeneous, for subsequent treatment has increased difficulty.Therefore how to make imager can have and be made for the key point that under different brightness conditions, obtains high quality graphic with the adaptive camera that is adjusted to of brightness.

Emerge in an endless stream at the imageing sensor to imaging under the low-light (level) environment on the hardware at present, this type transducer is to improve imaging effect through the sensitivity that improves the CCD element basically.The development of domestic low-light (level) imageing sensor imager mainly divides three phases: colour/black and white suitching type, low no-delay gate formula with telegraphy degree formula.Though these three kinds of cameras have solved low light to a certain extent according to the imaging problem under the condition; But have two big shortcomings: a shortcoming is to involve great expense, and the low-light (level) camera of better performances adopts the object under the infrared lens induction half-light usually, but this will increase the cost of camera greatly; The another one shortcoming is that function is comparatively fixing; Lack flexibility, promptly the low-light (level) index of this type imageing sensor receives the qualification of CCD performance, does not have the ability of regulating with the illumination self adaptation.Adopt infrared lens low-light (level) imageing sensor under the normal illumination condition, to influence the definition of image owing to the infrared composition in the natural daylight; Therefore all there is certain limitation in the low-light-level imaging appearance on range of application.

On software, mainly contain traditional overall enhancement algorithms and local enhancement methods to the algorithm that improves the low-light (level) picture contrast.Overall situation enhancement algorithms mainly contains histogram equalization and contrast enhancement algorithms etc.Histogram equalizing method is the method that increases contrast indirectly, and this method is simple, and effect is obvious, but exists enhancing back image shortcomings such as details disappearance to occur.Thereby contrast enhancement process is the distance between the direct stretched pixels strengthens contrast, for the different zone of image, and the function that local detail can not self adaptation strengthens.In processing procedure, be regardless of simultaneously noise and detailed information, when strengthening image gray levels difference, make noise also be enhanced.Overall in addition enhancement algorithms is not suitable for the bigger situation of the even contrast variation's amplitude of uneven illumination at night.The method of partial gradient domain model is a kind of local method that improves contrast; The dynamic range that these class methods have effectively been compressed the uneven illumination image has increased the contrast of dark place simultaneously; But the detail section reinforced effects to brightness is higher is not obvious; These class methods are bigger than the Global Algorithm amount of calculation simultaneously, and computation complexity is high.

Three, summary of the invention

The objective of the invention is to the high shortcoming such as popularization that is difficult to of illumination condition limitation, cost to traditional low-light-level imaging appearance; The imitation compound eye changes the process of regulating neural amalgamation mode with external environment illumination brightness; Cartridge fusion mechanism with the overlapping type compound eye of nerve is the basis, but adopts the multi-charge coupled apparatus assembly of imitative compound eye curved-surface structure to close the formation method of continuous monitoring round the clock that the DSP+FPGA chip has made up a kind of self-adapting adjustment type of cheap whole day monitoring.

Adopt following imager:

Comprise 3 identical CCD cameras, 3 are used for first kind DSP digital signal processing chip that luminance component separates with two color difference components, 1 and are used to accomplish second type of DSP digital signal processing chip, 3 programmable vision signal processing chips that are used for analog-to-digital conversion and system conversion, 3 electricity that multiple image merges and can wipe storage chip;

Above-mentioned 3 CCD closely link together, and being attached to a radius is on the curved surface of r, and the angle between adjacent two ommatidium optical axises is the curved-surface structure that θ ≈ artsin (D/r) constitutes imitative compound eye, and wherein r is the eye radius of compound eye, and D is the ommatidium diameter;

Above-mentioned 3 CCD cameras carry out preliminary treatment through the input end of analog signal that USB interface is connected to above-mentioned 3 programmable vision signal processing chips respectively; The state information of programmable vision signal processing chip is exported in the complex programmable logic device (CLPC) as control signal; Data-signal then can be wiped storage chip through above-mentioned 3 electricity and be input to respectively in above-mentioned 3 first kind DSP dsp core, and the control signal of 3 first kind DSP digital signal processing chips is input to 1 second type of DSP digital signal processing chip as chip selection signal and enable signal.

Formation method may further comprise the steps:

(1), utilizing three ommatidiums in the CCD cameras simulation compound eye to distinguish independently forms images, obtains three width of cloth ommatidium images;

(2), utilize the programmable vision signal processing chip respectively three secondary ommatidium images processes to be carried out the conversion of analog-to-digital conversion and standard;

(3), utilize three first kind DSP digital signal processing chips that three width of cloth ommatidium images are carried out separating of luminance component and two color difference components respectively;

(4), according to the characteristics of only luminance component being handled in the visual processes; Extract the Y component of three secondary ommatidium images; Color difference components together with not doing any processing after separating is input to the multiple image fusion treatment that inspired by the biological compound eye cartridge jointly, and wherein cartridge refers to a pixel of the corresponding fusion image of each cartridge; Locational three sample values of respective pixel get into same cartridge in three ommatidiums, and concrete blending algorithm is following:

Step 1: establish I1, I2 and I3 and represent three secondary ommatidium images respectively, extract the luminance component of three width of cloth ommatidium images; With any one secondary ommatidium image I 1 image as a reference, I2 and I3 are registrated on the reference picture I1;

Step 2: calculate the mean value of cartridge interoceptor, i.e. average to a plurality of CCD sampled values on the same position in the space; Calculate the average that centers on the local field interior pixel of a point in the space simultaneously;

Step 3: according to the comparison of cartridge interoceptor average and cartridge average; Select suitable syncretizing mechanism; When the brightness average around the cartridge is higher than the mean value of cartridge interior lights receptor; The extraneous brightness of explanation is lower so, and the information that need three photoreceptors be obtained is through reinforcements that superpose of receptor cell wall, and raising is to the sensitiveness of brightness; Otherwise explain that this ambient light degree of illuminating is higher, need cartridge be belonged to the image of the contrast raising acquisition enhancing in field so;

Suppose the ommatidium image I that CCD1 obtains ₁Image as a reference; I _k(l) expression k width of cloth image is arranged the brightness of l the pixel in back according to the column vector form; F _Mean(l), L _Mean(l) be the average and reference picture I of photoreceptor brightness in l the cartridge respectively ₁Going up with l cartridge is the mean flow rate of 3 * 3 cartridges at center; Owing to have only three ommatidiums, so the number of the photoreceptor in each cartridge is 3:

$F_{\mathrm{mean}}\left(l\right)=\underset{k=\mathrm{1,2,3}}{\mathrm{mean}}\left(I_k\left(l\right)\right);L_{\mathrm{mean}}\left(l\right)=\underset{k=l-3,l+3}{\mathrm{mean}}\left(L\left(k\right)\right);---\left(1\right)$

Work as F in the formula _Mean(l)＞L _Mean(l) intensity of illumination in the explanation external world is higher the time, adopts first kind of amalgamation mode, otherwise adopts second kind of amalgamation mode; Before merging, image is arranged by the form of column vector;

(1), shunting strengthen be under the normal illumination condition to the same scene position on the fusion treatment of different information; The stimulation that k photoreceptor of l cartridge accepted is I _kWhat (l), k receptor moderate stimulation was the strongest is

What stimulation was the most weak is

Mean flow rate

So owing to be stronger illumination condition F _Mean(l)＞L _Mean(l), the therefore extracellular space V of l cartridge _Ecs(l) will receive input, and deduct the electric current of the photoreceptor of the more weak stimulation of the flow direction than the high light receptor; The voltage of l cartridge extracellular space induction generation is v _Ecs(l):

$V_{\mathrm{ecs}}\left(l\right)=L_{\mathrm{mean}}\left(l\right)+a_1\×(e^{\underset{k=\mathrm{1,2,3}}{\mathrm{\Σ}}(I_{\max }\left(l\right)-I_k\left(l\right))}-e^{\underset{k=\mathrm{1,2,3}}{\mathrm{\Σ}}(I_{\mathrm{mean}}\left(l\right)-I_{\min }\left(l\right))})---\left(2\right)$

A wherein ₁, be parameter, L _Mean(l) be the brightness average of 3 * 3 cartridges on the reference picture that is the center with l cartridge; Here the mediator exponentially that adopts exponential function to discharge when explaining that brightness is higher increases, and when stimulus of receptor hour, the mediator of release will approach 0, even has backward current and produce;

(2), the stack loop is under the not high situation of intensity of illumination, when the mean value of the monochrome information that receives when receptor is lower than the average of adjacent cartridge, explain extraneous brightness low be F _Mean(l)＜L _Mean(l), then open stack loop fusion mode; Here adopt voltage L with adjacent cartridge _Mean(l) with l cartridge internal interface receive brightness value I _kThe absolute value of difference (l) improves the brightness of receptor, so the voltage V of extracellular space _Ecs(l) be:

$V_{\mathrm{ecs}}\left(l\right)=L_{\mathrm{mean}}\left(l\right)+a_2\×\underset{k=\mathrm{1,2,3}}{\mathrm{\Σ}}{e}^{|I_k\left(l\right)-L_{\mathrm{mean}}\left(l\right)|}---\left(3\right)$

The voltage of each ammunition socket joint is L when initial _Mean(l), calculate,, influence v thereby receiver can produce response according to the size that stimulates along with the increase of outside stimulus through formula (1) _Ecs(l); A in formula (2) and (3) ₁, a ₂Coefficient is different, and one is the contrast that on the basis of mean flow rate, improves image, and another one is on the basis of the original brightness of background, to increase brightness; Coefficient a ₁, a ₂All be specific conductance, span (0,1] between, a wherein ₁Be the specific conductance between the adjacent receptor (pixel), and a ₂Be the specific conductance between cartridge and the receptor, because the absolute construction of cartridge, its inner electric current flows frequent, and outside when bigger stimulation does not occur, and does not have the flow of current process basically.Therefore can estimate a according to the structure of cartridge ₁＞a ₂Here select a ₁=0.5; a ₂=0.25;

Step 4: under the effect of the inner extracellular voltage of cartridge, obtain the output valve that a cartridge merges at last; The extracellular space voltage v of a cartridge _Ecs(l) meeting receives cartridge extracellular space z on every side _l-z _{L '}Influence, so cartridge external space voltage strength is expressed as:

$z_l=V_{\mathrm{ecs}}\left(l\right)+c\×\underset{l-l^{\′}=1}{\mathrm{\Σ}}(L_{\mathrm{mean}}\left(l\right)-L_{\mathrm{mean}}\left(l^{\′}\right))---\left(4\right)$

First v in this formula behind the equal sign _Ecs(l) inherent voltage of expression cartridge interoceptor, second L _Mean(l)-L _MeanThe difference of brightness between l cartridge of (l ') expression and the individual cartridge of l ', the difference of the respective pixel between the promptly adjacent ommatidium image, the size of this difference have influenced the size of cartridge external space voltage; Coefficient c is the coefficient of conductivity of extracellular voltage and cartridge; Also be a kind of specific conductance, span (0,1] between; L ' has represented other cartridges except that l cartridge; Distance between the cartridge has determined the intensity that influences between the cartridge, and obviously along with the increase of distance, influence each other will reduce the impact effect of the cartridge of close together; Here chosen distance cartridge l is that 1 pixel distance is with interior cartridge; But the c conducting is adjacent cartridge, because under common state, separate between cartridge and the cartridge, cynapse does not connect, but the ion flow of cytoplasm is conducted simultaneously, so therefore specific conductance is less, and c=0.1;

According to cartridge external space voltage z _lCalculate the result of l cartridge after having merged several ommatidium images for I ' (l):

I′(l)＝q×z _l+I ₁(l) (5)

Wherein I ' is to have obtained last output result, I after l cartridge merged three receptors (l) ₁(l) be current with reference to ommatidium figure I ₁Last l pixel value; Q is the neuronic cynapse connection weight of cartridge and second order; It is mobile more fast than flooding mechanism that the connection of passing through cynapse like this produces electric current, so the coefficient of conductivity is wanted select q=10 here greatly; After neural fusion treatment finishes I ' is passed through the image that the YUV inverse transformation obtains colored output with colouring information.

The cartridge model is according to a kind of Artificial Neural Network of Moya proposition in 1998.

List of references:

[1]Haines?K.G.，Moya?J.A.，Caudell?T.P.Modeling?Nonsynaptic?Communication?Between?Neurons?in?the?Lamina?Ganglionaris?ofMusca?Domestica[C].1999：

[2]Moya?J.A.A?comprehensive?model?ofthe?neuro-ommatidial?layer?in?Musca[D].University?of?New?Mexico，1998.

[3] Haines K.G., Pearlmutter B.A., Moya J.A., wait .Visualizing Communication Between Neurons in the Lamina Ganglionaris ofMusca domestica [C] .2000:

This neural net is a kind of feedforward network of describing single cartridge inside and cartridge connection on every side thereof.Described from amphiblestroid light receiving unit to the information conduction pattern the high-order neuron LMC cell.This neural net comprises three processing: the coupling between inner six photoreceptors of same cartridge; The ions diffusion of cell colloid and cartridge and the neural LMC cell of second order is connected between cartridge and the cartridge.Receptor receives extraneous brightness to stimulate, and the difference of stimulation has caused producing flow of current between inner six receptors of a cartridge.The cartridge volume inside exist can conduction current charged ion, the electromotive force of photoreceptor can carry out the diffusion of electric current through such transmission medium.The neuronic connection of cartridge and second order is that a kind of cynapse connects, and can the high-order neuron that environmental stimuli is delivered to brain be gone to handle.This neural net can be regarded the UNE of the single output of a kind of many inputs as; Input has obtained the input of second order nerve cell LMC cell in the time of a plurality of ommatidium image, and this fusion process can be extracted high-frequency information, removes redundant information.

The description of three following these neural nets of formula:

$\frac{{\∂p}_{k,i}}{\∂t}=-p_{\mathrm{ki}}{A}^{\′}+(B^{\′}-p_{\mathrm{ki}}{C}^{\′})I_{k,i}^{+}+(p_L-2p_{\mathrm{ki}}+p_R)D^{\′}+(Z_k-p_{\mathrm{ki}})E---\left(6\right)$

$\frac{{\∂z}_k}{\∂t}=-\underset{i}{\mathrm{\Σ}}(p_{k,i}-z_k)E-\underset{l}{\mathrm{\Σ}}(z_k-z_j)L---\left(7\right)$

$\frac{{\∂a}_{L,i}}{\∂t}=-a_{L,j}F-\underset{i=\mathrm{1,6}}{\mathrm{\Σ}}{G}_i(p_i-z)---\left(8\right)$

A in this model, B, C, D, E, F and G are coefficients.They are respectively:

A＝0.7；B＝1.0；C＝1.0；D＝0.25；E＝0.1；F＝3.0；G＝10.0；L＝30.0。

The principle explanation:

Mostly imaging device and equipment based on the insect compound eye structural is its biological structure characteristics of imitation at present, so that obtain panoramic picture.And the neural processor system of compound eye also only is widely used in target travel detects.About adopting compound eye self adaptive imaging mechanism to be implemented in the imaging device that obtains high quality graphic under the different illumination conditions, do not appear in the newspapers.

In fact; The characteristic of the multiple aperture of compound eye is corresponding, and a whole set of is different from the nerve regulation mechanism and the syncretizing mechanism of single aperture eye; Can be adaptive when illumination condition changes the information fusion mode of adjusting intrinsic nerve unit, make insect obtain clear reliable scene information in the different illumination condition.This visual characteristic of compound eye can be explained with the mechanism that a plurality of ommatidium images in the compound eye thin plate layer cartridge merge.What each cartridge was collected is the photoreceptor information from different ommatidiums, the repeatedly sampled value of a point in the extraneous scene that they are corresponding.Through the electric coupling relationship together, coupled modes change along with the change of intensity of illumination between the photoreceptor.When illumination condition is more excellent; Coupling between the photoreceptor is the process that a kind of contrast stretches; Thereby receive than strong stimulation and produce the photoreceptor cell that electric current is diverted to more weak stimulation than the receptor of big electric current; Discharge chemical mediator, the balance depolarising of breaking external space ion concentration to extracellular space simultaneously.And extracellular space can be lower than the receptor generation adverse current of external space inherent voltage to aixs cylinder voltage.Realized like this under optimum illumination condition, externally the effect of the equalization of scene stimulation and contrast lifting.When light application ratio was dark, the electric current between the photoreceptor was exactly a kind of mode that superposes and merge.The inherent voltage of extracellular space is lower, is not high enough to the generation backward current, and the electric current of a plurality of photoreceptors improves the intensity of receptor to external space output current through the coupling stack.Extracellular space has been transmitted extraneous stimulus information as a kind of intermediate medium to more higher leveled nerve cell through non-cynapse connection, and the contrast of image and the sensitiveness of brightness are regulated in adaptive brightness according to image background in transmittance process.

It is thus clear that compound eye has the treatment mechanism that improves picture quality when merging the ommatidium image, main characteristics reduces: 1. adopt Local treatment mechanism, need not to calculate, the background luminance through the part feeds back to photoreceptor and controls contrast.The contrast of the image that the electric current shunting between the photoreceptor and the fusion of extracellular space voltage improve automatically, the adaptive electric current amalgamation mode that changes photoreceptor according to local luminance.2. the scope of the variation of the electric current in the cartridge is higher than the brightness excursion far away, and bigger dynamic range can reflect image information more accurately.3. because photoreceptor has lower sensitiveness,, make in the raising of contrast, to have stronger antijamming capability so do not receive saturated or The noise.

Of the present invention having the following advantages: thus curved surface multi-charge coupled apparatus group self-adaptive imaging instrument imitation compound eye obtains the principle of self adaptive imaging effect with the change adjustment nerve conduction mode of illumination condition under day alternates with night situation, adopt a plurality of common charge-coupled device (CCD)s and corresponding auxiliary circuit and neural processing algorithm to make up a kind of bionical self-adapting adjustment type imager.Can be according to the different illumination condition; Adaptive switching to obtained accurate, distinct image information on the pictures different blending algorithm; Broken through traditional low-light (level) camera relatively poor limitation of imaging effect by day; It not only can obtain the image of high-contrast under the normal illumination condition, can also under the weak light condition, obtain the image of high brightness sensitiveness, the monitoring that realized that day alternates with night.The blending algorithm of this imager is simple, and amount of calculation is little, can requirement of real time.

Four, description of drawings

The arrangement mode of accompanying drawing 1:CCD camera;

Accompanying drawing 2: the logic connection layout of circuit of the present invention;

Accompanying drawing 3: flow chart of the present invention;

Accompanying drawing 4: shunting enforcement module flow process;

Accompanying drawing 5: stack loop module flow process;

Accompanying drawing 6:E2PROM and DSP connection layout;

Five, embodiment

The present invention has designed the multi-charge coupled apparatus group self-adaptive imaging instrument of curved face type optical texture.Main hardware device has 3 common colored CCD cameras, and two DSP digital signal processing chip TMS320DM642 and TMS320VC5510, programmable vision signal processor SAA7111A and electricity with powerful quick operational capability can be wiped storage chip E2PROM.

The model of CCD is 3209A, and this camera has two kinds of standards: Phase Alternation Line system and NTSC, resolution is respectively 500 * 582,510 * 429 two kinds.Adopt pal mode among the present invention, camera lens is of a size of 38mm * 38mm.The resolution of the ommatidium image that obtains is 500 * 582.Shown in accompanying drawing 1,3 CCD are bundled closely, 3 ommatidiums on the curved surface that to be attached to a radius be r in the simulation insect compound eye, the angle between every adjacent two ommatidium optical axises is θ ≈ artsin (D/r).Constitute the curved-surface structure of imitative compound eye.As shown in Figure 3, the o point is the central point of compound eye, and r is the eye radius of compound eye, and D is ommatidium diameter 38mm.

In the dsp chip model: a kind of is TMS320DM642, is aly in the C6000 of the TI company series 32 to be dsp chip to what the multimedia process field was used, and its core is a C6416 type high-performance digital signal processor; Have extremely strong handling property, the flexibility and the programmability of height, peripheral simultaneously integrated: three configurable video interfaces; Can import with video; Control port (VIC), I are inserted in output or MPTS input seamless link in the VCXO ²Equipment and interfaces such as the audio frequency that C bus module etc. are very complete, video and network service are specially adapted to the high-speed dsp application based on digital video/image processing.A kind of in addition is TMS320VC5510; This chip is 16 high fixed-point dsps of a cost performance of releasing in 2000; Adopt advanced correction Harvard structure; 8 buses, CPU are arranged in the sheet, at the sheet memory with at hardware such as sheet peripheral circuits, add the highly command system of specialty, make it have advantages such as power consumption is little, highly-parallel.

Also have video signal preprocessor SAA7111A in addition, this chip is accomplished the conversion of video formats, and the coloured image of yuv format is provided.Push-up storage (FIFO) IDT72V2113 in addition that cooperates dsp chip to use, its maximum operating frequency is 133MHz; Capacity is 512KB.CPLD (CPLD) LC4128V is that memory and dsp chip provide logic control signal.The E2PROM of AT28C256 model is the electricity erasable memorizer of 32k * 8.

The method of attachment of hardware is shown in accompanying drawing 2; This curved-surface structure obtains three ommatidium images simultaneously; Each CCD colour imagery shot is input to the digital-to-analogue conversion that vision signal conversion chip SAA7111A go up to accomplish signal through USB interface with analog signal, the amplification of signal and preliminary treatment such as anti-aliasing.SAA7111A output state information and data message, it is last as control signal that state information output is sent to complex programmable logic device (CLPC) LC4128V; Data message is being input on the dsp chip TMS320VC5510 under the CPLC control signal after then cushioning through first in first out storage chip (FIFO) IDT72V2113.

Three TMS320VC5510 chips are accomplished the chrominance information of three way word color picture signals and the separating treatment of monochrome information respectively.According to the characteristics of only luminance component being handled in the visual processes, three dsp chip TMS320VC5510 go up the Y component are extracted respectively, are input on the next stage TMS320CM642 type dsp chip and carry out fusion treatment, and be left intact after the color difference components input.

TMS320VC5510 chip and next stage dsp chip TMS320DM642 are the annexations of master-slave mode, and three TMS320VC5510 are from sheet, and TMS320DM642 is as main leaf.Main leaf adopts interrupt mode that three dateouts from sheet are read.Each TMS320VC5510 from sheet all to main leaf input data information and state information.What broad arrow was represented is data message, the state information of thin arrow representative output.State information is the interrupt signal of sending to main leaf from sheet DSP, and the procedure turn of interrupting the master chip operation is read from the data of chip output.Main leaf TMS320DM642 goes up and accomplishes neural fusion treatment and the monochrome information of colour information after fusion treatment obtained colored output image through the YUV inverse transformation.

Fig. 3 is the flow chart of this patent.Mainly contain three big modules: acquisition module, pretreatment module, Fusion Module.

Acquisition module cooperates the vision signal conversion chip to realize respectively through three CCD.The CCD camera of 3 same physical parameter is arranged arrangement according to the curved surface of imitative compound eye structural.The image that each CCD camera obtains is a width of cloth ommatidium image, and each pixel in the ommatidium image is a photoreceptor in the ommatidium, and the resolution of the CCD that selects here is 500 * 582.

Pretreatment module is respectively the monochrome information of the digital picture that obtains separately to be separated with colour difference information through three dsp chips.The monochrome information of separating and chrominance information are transferred to next stage respectively through data wire main dsp chip.This module guarantees importing into synchronously of three width of cloth ommatidium images that synchronization is taken also under the control of CPLC.

Fusion Module is the core of this patent, and the TMS320DM642 core is realized on main leaf, and this master chip has also connected two E2PROM storage chip AT28C256, is used for storing data and program.This module receives from monochrome information and chrominance information after three dsp chips processing, only monochrome information is carried out registration, finds locational three sample values of respective pixel in three ommatidiums that will get into same cartridge.A pixel on the corresponding fusion image of each cartridge.Because the angle between the known image in design process can pass through direct registration three width of cloth images of affine transformation.Select any pair to be reference picture, other two width of cloth are registrated on this reference picture.The formula of affine transformation:

$\left(\begin{array}{c}x\\ y\end{array}\right)=\left(\begin{array}{cc}{a}_{11}& a_{12}\\ a_{21}& a_{22}\end{array}\right)\left(\begin{array}{c}{x}^{\&prime;}\\ y^{\&prime;}\end{array}\right)+\left(\begin{array}{c}{b}_1\\ {\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="H2009102641659E00031.png"\; state="\{\{state\}\}">b</figure-callout>}}_2\end{array}\right)$

Wherein: element representation rotation and zoom factor in the A matrix, b vector representation shift factor.Here select I ₁Image adopts two matrix A as a reference ₁, A ₂I is described ₂, I ₃With respect to I ₁Twiddle factor.The cartridge structure need not judged illumination condition on the overall situation, but decides which kind of amalgamation mode of employing through the analysis of part.First kind of amalgamation mode is that shunting strengthens, and another kind of amalgamation mode is the stack loop.

When ambient light degree of illuminating is higher; Current imbalance between three photoreceptors; Can produce coupling phenomenon between the adjacent in twos receptor; Stimulate bigger receptor will shunt a part to stimulating less receptor, the voltage of the relative extracellular space of receptor that each acceptance simultaneously stimulates is higher, and the release that produces chemical mediator at the aixs cylinder position realizes depolarising.If the stimulation that certain receptor is accepted is very little, the voltage of generation is far smaller than the inherent voltage of extracellular space, will produce backward current so.When therefore in stimulation, having sudden change brightness, different information flows between a plurality of photoreceptors, has shared the phenomenon that single receptor is accepted the big electric current of undue intense stimulus generation.Such shunting simultaneously makes the contrast of image improve automatically.Shown in the accompanying drawing 4, each photoreceptor node is all told certain electric current to two adjacent receptors, photoperceptive output that the difference between three environmental stimulis is corresponding, thus produced the induced current of extracellular space.Such coupling mechanism just shows a kind of inhibition of sudden change brightness.

Opposite lower when extraneous brightness, the brightness that the inner receptor of cartridge receives causes difference hour because extraneous brightness is lower, superpose between the receptor with the sensitiveness of increase to light.Shown in accompanying drawing 5, information circulation between three nodes is strengthened, and the induced current of extracellular space is that the difference of environmental stimuli and external space average voltage intensity produces.

The hardware connection layout is as shown in Figure 6, has shown the connected mode of storage chip E2PROM and dsp chip among Fig. 7, has deposited the program and the data of main leaf operation in this dsp chip.The CEn end of dsp chip is connected through a not gate with the section of enabling end of E2PROM, and low level is effective.Output enable end AOE end is held with

and is linked to each other through a not gate, and low level is effective.The Enable Pin AWE that writes of DSP links to each other through a not gate with

end of storage chip.The address is to be outputed on A [15...0] end of storage chip by EAxx, and data terminal ED [15...0] links to each other with the input/output port of storage chip.

The present invention copies the mechanism of compound eye adaptive adjusting amalgamation mode with the change of illumination condition; By the auxiliary multi-charge coupled apparatus group self-adaptive imaging instrument of having realized of common imaging device and many dsp chips group and some storage chips and logic control chip, realize obtaining of round-the-clock high quality graphic.There is overlapping image in three width of cloth that adopt three common CCD cameras that are fixed on the arch face to obtain Same Scene.These superimposed images carry out convergence analysis on chip TMS320DM642, provide the illumination condition discrimination formula according to present image information, select suitable amalgamation mode to obtain last output at last.

Advantage of the present invention: the shortcoming of traditional low-light-level imaging appearance high cost and complexity has been avoided in this invention design; Make full use of biological vision simply fast analysis principle cooperate flexibly nerve conduction mechanism simultaneously, combination comes fast with fpga chip to rely on DSP, the replacement PC of dexterity has been realized bionical neural fusion treatment; The adaptive scene image that obtains; Because algorithm is simple, requirement of real time, and have advantage such as popularization of being easy to.

Claims (1)

1. the multi-charge coupled apparatus group self adaptive imaging method of a curved surface optical texture is characterized in that:

Adopt following imager:

Comprise 3 identical CCD cameras, 3 are used for first kind DSP digital signal processing chip that luminance component separates with two color difference components, 1 and are used to accomplish second type of DSP digital signal processing chip, 3 programmable vision signal processing chips that are used for analog-to-digital conversion and system conversion, 3 electricity that multiple image merges and can wipe storage chip;

Above-mentioned 3 CCD closely link together, and being attached to a radius is on the curved surface of r, and the angle between adjacent two ommatidium optical axises is the curved-surface structure that θ ≈ artsin (D/r) constitutes imitative compound eye, and wherein r is the eye radius of compound eye, and D is the ommatidium diameter;

Above-mentioned 3 CCD cameras carry out preliminary treatment through the input end of analog signal that USB interface is connected to above-mentioned 3 programmable vision signal processing chips respectively; The state information of programmable vision signal processing chip is exported in the complex programmable logic device (CLPC) as control signal; Data-signal then can be wiped storage chip through above-mentioned 3 electricity and be input to respectively in above-mentioned 3 first kind DSP dsp core, and the control signal of 3 first kind DSP digital signal processing chips is input to 1 second type of DSP digital signal processing chip as chip selection signal and enable signal;

Formation method may further comprise the steps:

(1), utilizing three ommatidiums in the CCD cameras simulation compound eye to distinguish independently forms images, obtains three width of cloth ommatidium images;

(2), utilize the programmable vision signal processing chip respectively three secondary ommatidium images processes to be carried out the conversion of analog-to-digital conversion and standard;

(3), utilize three first kind DSP digital signal processing chips that three width of cloth ommatidium images are carried out separating of luminance component and two color difference components respectively;

(4), according to the characteristics of only luminance component being handled in the visual processes; Extract the Y component of three secondary ommatidium images; Color difference components together with not doing any processing after separating is input to the multiple image fusion treatment that inspired by the biological compound eye cartridge jointly, and wherein cartridge refers to a pixel of the corresponding fusion image of each cartridge; Locational three sample values of respective pixel get into same cartridge in three ommatidiums, and concrete blending algorithm is following:

Step 1: establish I1, I2 and I3 and represent three secondary ommatidium images respectively, extract the luminance component of three width of cloth ommatidium images; With any one secondary ommatidium image I 1 image as a reference, I2 and I3 are registrated on the reference picture I1;

Step 2: calculate the mean value of cartridge interoceptor, i.e. average to a plurality of CCD sampled values on the same position in the space; Calculate the average that centers on the local field interior pixel of a point in the space simultaneously;

Step 3: according to the comparison of cartridge interoceptor average and cartridge average; Select suitable syncretizing mechanism; When the brightness average around the cartridge is higher than the mean value of cartridge interior lights receptor; The extraneous brightness of explanation is lower so, and the information that need three photoreceptors be obtained is through reinforcements that superpose of receptor cell wall, and raising is to the sensitiveness of brightness; Otherwise explain that this ambient light degree of illuminating is higher, need cartridge be belonged to the image of the contrast raising acquisition enhancing in field so;

Suppose the ommatidium image I that CCD1 obtains ₁Image as a reference; I _k(l) expression k width of cloth image is arranged the brightness of l the pixel in back according to the column vector form; F _Mean(l), L _Mean(l) be the average and reference picture I of photoreceptor brightness in l the cartridge respectively ₁Going up with l cartridge is the mean flow rate of 3 * 3 cartridges at center; Owing to have only three ommatidiums, so the number of the photoreceptor in each cartridge is 3:

$F_{\mathrm{mean}}\left(l\right)=\underset{k=\mathrm{1,2,3}}{\mathrm{mean}}\left(I_k\left(l\right)\right);L_{\mathrm{mean}}\left(l\right)=\underset{k=l-3,l+3}{\mathrm{mean}}\left(L\left(k\right)\right);---\left(1\right)$

Work as F in the formula _Mean(l)＞L _Mean(l) intensity of illumination in the explanation external world is higher the time, adopts first kind of amalgamation mode, otherwise adopts second kind of amalgamation mode; Before merging, image is arranged by the form of column vector;

(1), shunting strengthen be under the normal illumination condition to the same scene position on the fusion treatment of different information; The stimulation that k photoreceptor of l cartridge accepted is I _kWhat (l), k receptor moderate stimulation was the strongest is

What stimulation was the most weak is

Mean flow rate

So owing to be stronger illumination condition F _Mean(l)＞L _Mean(l), the therefore extracellular space V of l cartridge _Ecs(l) will receive input, and deduct the electric current of the photoreceptor of the more weak stimulation of the flow direction than the high light receptor; The voltage of l cartridge extracellular space induction generation is v _Ecs(l):

$V_{\mathrm{ecs}}\left(l\right)=L_{\mathrm{mean}}\left(l\right)+a_1\&times;(e^{\underset{k=\mathrm{1,2,3}}{\mathrm{\&Sigma;}}(I_{\max }\left(l\right)-I_k\left(l\right))}-e^{\underset{k=\mathrm{1,2,3}}{\mathrm{\&Sigma;}}(I_{\mathrm{mean}}\left(l\right)-I_{\min }\left(l\right))})---\left(2\right)$

A wherein ₁Be parameter, L _Mean(l) be the brightness average of 3 * 3 cartridges on the reference picture that is the center with l cartridge; Here the mediator exponentially that adopts exponential function to discharge when explaining that brightness is higher increases, and when stimulus of receptor hour, the mediator of release will approach 0, even has backward current and produce;

(2), the stack loop is under the not high situation of intensity of illumination, when the mean value of the monochrome information that receives when receptor is lower than the average of adjacent cartridge, explain extraneous brightness low be F _Mean(l)＜L _Mean(l), then open stack loop fusion mode; Here adopt voltage L with adjacent cartridge _Mean(l) with l cartridge internal interface receive brightness value I _kThe absolute value of difference (l) improves the brightness of receptor, so the voltage V of extracellular space _Ecs(l) be:

$V_{\mathrm{ecs}}\left(l\right)=L_{\mathrm{mean}}\left(l\right)+a_2\&times;\underset{k=\mathrm{1,2,3}}{\mathrm{\&Sigma;}}{e}^{|I_k\left(l\right)-L_{\mathrm{mean}}\left(l\right)|}---\left(3\right)$

The voltage of each ammunition socket joint is L when initial _Mean(l), calculate,, influence v thereby receiver can produce response according to the size that stimulates along with the increase of outside stimulus through formula (1) _Ecs(l); A in formula (2) and (3) ₁, a ₂Coefficient is different, and one is the contrast that on the basis of mean flow rate, improves image, and another one is on the basis of the original brightness of background, to increase brightness; Coefficient a ₁, a ₂All be specific conductance, span (0,1] between, a wherein ₁Be the specific conductance between the adjacent receptor, and a ₂Be the specific conductance between cartridge and the receptor, because the absolute construction of cartridge, its inner electric current flows frequent, and outside when bigger stimulation does not occur, and does not have the flow of current process basically, therefore can estimate a according to the structure of cartridge ₁＞a ₂Here select a ₁=0.5; a ₂=0.25;

Step 4: under the effect of the inner extracellular voltage of cartridge, obtain the output valve that a cartridge merges at last; The extracellular space voltage v of a cartridge _Ecs(l) can receive cartridge extracellular space z on every side _l-z _{L '}Influence, so cartridge external space voltage strength is expressed as:

$z_l=V_{\mathrm{ecs}}\left(l\right)+c\&times;\underset{l-l^{\&prime;}=1}{\mathrm{\&Sigma;}}(L_{\mathrm{mean}}\left(l\right)-L_{\mathrm{mean}}\left(l^{\&prime;}\right))---\left(4\right)$

First v in this formula behind the equal sign _Ecs(l) inherent voltage of expression cartridge interoceptor, second L _Mean(l)-L _MeanThe difference of brightness between l cartridge of (l ') expression and the individual cartridge of l ', the difference of the respective pixel between the promptly adjacent ommatidium image, the size of this difference have influenced the size of cartridge external space voltage; Coefficient c is the coefficient of conductivity of extracellular voltage and cartridge; Also be a kind of specific conductance, span (0,1] between; L ' has represented other cartridges except that l cartridge; Distance between the cartridge has determined the intensity that influences between the cartridge, and obviously along with the increase of distance, influence each other will reduce the impact effect of the cartridge of close together; Here chosen distance cartridge l is that 1 pixel distance is with interior cartridge; But the c conducting is adjacent cartridge, because under common state, separate between cartridge and the cartridge, cynapse does not connect, but conducts through the ion flow of cytoplasm, so specific conductance is less, c=0.1;

According to cartridge external space voltage z _lCalculate the result of l cartridge after having merged several ommatidium images for I ' (l):

I′(l)＝q×z _l+I ₁(l) (5)

Wherein I ' is to have obtained last output result, I after l cartridge merged three receptors (l) ₁(l) be current with reference to ommatidium figure I ₁Last l pixel value; Q is the neuronic cynapse connection weight of cartridge and second order; It is mobile more fast than flooding mechanism that the connection of passing through cynapse like this produces electric current, so the coefficient of conductivity is wanted select q=10 here greatly; After neural fusion treatment finishes I ' is passed through the image that the YUV inverse transformation obtains colored output with colouring information.

Images