CN103489312A - Traffic flow information collection method based on image compression - Google Patents

Abstract

The invention discloses a traffic flow information collection method based on image compression. The method comprises the steps of conducting moving object detection on collected videos through the background subtraction, conducting preprocessing, such as filtering and noise removing, on a digital traffic image, utilizing an improved embedded zerotree wavelet encryption algorithm for conducting the image compression, conducting data transmission on a compressed code stream, conducting decompression to obtain a reconstructed image, and conducting vehicle statistics and calculating traffic data according to the identified target information. The traffic flow information collection method comprises the following steps: collecting the video image, detecting a moving object, preprocessing the traffic image, compressing the traffic image, transmitting the compressed code stream, conducting the decompression, conducting the vehicle statistic, and calculating the traffic flow information data. The improved embedded zerotree wavelet encryption algorithm enables the scanning precision to be increased, the number of the data is reduced, and the requirement of the intelligent traffic field for storing and transmitting information collection images at signalized intersections can be well met.

Description

A kind of telecommunication flow information acquisition method based on compression of images

Technical field

The invention belongs to intelligent transportation field, be specifically related to a kind ofly based on the application of improved Embedded Zerotree wavelet Image Compression Algorithm in intelligent transportation, realize the real-time monitoring of road traffic condition and obtain the traffic flow parameter data.

Background technology

Intelligent transportation system is by detecting in real time the parameter informations such as traffic flow of road road network, understand the operation conditions of road traffic, dynamic change according to the road traffic flow, make rapidly organization and administration and the optimal control of road traffic, alleviate the congestion in road degree, reduce the vehicle driving delay time at stop, guarantee traffic safety, reduce the probability of accident, and road traffic infrastructure is fully used, finally reach the purpose of intelligent transportation.Traffic information collection based on intelligent transportation system, changed traditional field measurement investigation, and the situation of the material resources that waste financial resources, by improving effectively making for improving the efficiency of traffic system traffic parameter information.

Monitoring and traffic information collection for the urban traffic road situation have different means, traffic flow data sampling based on video image is to utilize video and image processing techniques, road traffic condition is monitored in real time, and obtain the mode of traffic flow parameter data by respective algorithms, be a kind of novel traffic parameter collecting method occurred in recent years.Utilizing camera video to catch gathers and image processing and analyzing, carrying out the telecommunication flow information data acquisition obtains, and the vehicle target in sequence of video images is detected and vehicle identification, extract the transport information parameter by the image recognition to vehicle and vehicle flowrate, complete the telecommunication flow information data acquisition based on video means.

Traffic parameter extraction system based on video image is to consist of video camera, computing machine treatment technology, microprocessor or industrial computer etc., relate to a plurality of kens such as image processing, computer vision, pattern-recognition, signal processing and information fusion, it has the following advantages: machinery and equipment is easy for installation, do not there is destruction, simultaneously, can effectively utilize existing video equipment on network of highways, this will reduce expenses greatly; Area coverage is large, and a set of video parameter extraction equipment can detect several tracks simultaneously, and the amount of traffic information of obtaining is abundant, has traffic monitoring and traffic management function; For vehicle supervision department provides visual image.

And, along with the development of intelligent transport technology and mechanics of communication, people are also more and more higher to the requirement of the transmission quality of digital traffic image information and transmission speed.The Digital Transportation picture signal comprises huge quantity of information, and the restriction of channel width and storage space has brought very large difficulty to practical application, and therefore research is of great significance the intelligent compression coding technology tool of real-time information collection image.

Wavelet transformation is the most outstanding achievement of 20 applied mathematics circle in the end of the century, its essence is the time m-dimensional analysis method of signal, can information extraction from signal effectively, by calculation functions such as flexible and translations, function or signal are carried out to multiscale analysis, be described as " school microscop ".Wavelet transform is exactly, on the basis of selecting suitable Orthogonal Wavelets, image is carried out to Multiresolution Decomposition for the basic thought of Image Coding, resolves into the subimage of different spaces, different frequency, then subimage is carried out to coefficient coding.The coefficient image generated has the localization of concentration of energy, spatial frequency, the characteristics such as gregariousness of significant coefficient, and these characteristics that have just because of wavelet coefficient, it is very suitable for embedded encoded algorithm.

Compression of images is a kind of coding techniques of view data, its objective is original larger image try one's best few byte representation and transmission, and requires restored image that preferable quality is arranged.Utilize compression of images, can alleviate the burden of image storage and transmission, make high-quality image realize fast transport and process in real time on network.In fact, compression of images adopts various technology to reduce redundant data exactly, utilizes human visual system's physiology and the various characteristics of psychological characteristic and image information source to obtain higher ratio of compression simultaneously.Existing Standard of image compression is based on the orthogonal transformation of image, obtained the good compression effect than low compression ratio the time, reconstructed image quality is good, but along with the increase of ratio of compression, reconstructed image quality sharply descends, and can't meet low bit rate hypograph compression requirement.The present invention is based on Embedded Zerotree Wavelet Coding, by introducing dynamic threshold and increasing the sign type to improve compression performance, this algorithm can carry out efficient compression of images, meets preferably the requirement of intelligent transportation field to the storage of real-time information collection image and transmission.

Summary of the invention

It is research object that the video sequence of signalized intersections video monitoring collection is take in the present invention, carry out moving object detection by the video to gathering with background subtraction and obtain the Digital Transportation image, it is carried out to filtering and remove the pre-service such as noise, and utilize improved Embedded Zerotree Wavelet Coding to carry out compression of images, code stream after compression is carried out to data transmission, and in terminal, compressed bit stream is decompressed and obtains reconstructed image, and then carry out car statistics and calculate traffic information data according to the target information of identification.

The present invention installs video camera in traffic control system, and machinery and equipment is easy for installation, does not have destruction, easily safeguards, and can change at any time position and the size of surveyed area, and area coverage is large, and the amount of traffic information of obtaining is abundant, does not affect traffic.It is very large that video detects the quantity of information of obtaining, and by the high-quality information of extractible Digital Image Processing, the traffic in the video image scene carried out to the intelligent real-time monitoring of efficiently and accurately, catches movable information promptly and accurately.Improved Embedded Zerotree Wavelet Coding improves compression performance by introducing dynamic threshold and increasing the sign type, has realized the quick judgement of zero tree construction, and has completed the improvement quantized successively.From performance index and the packed data measuring angle of reconstructed image, improved Embedded Zerotree Wavelet Coding can meet the compression transmission demand of information acquisition image in the signalized intersections monitoring.

The technical scheme that realizes the object of the invention is: a kind of telecommunication flow information acquisition method based on compression of images, comprise that the video to gathering carries out moving object detection with background subtraction, digital traffic image is carried out to filtering and remove the pre-service such as noise, utilize improved Embedded Zerotree Wavelet Coding to carry out compression of images and the code stream after compression is carried out to data transmission and reconstructed image is obtained in decompression, carry out car statistics and calculate traffic information data according to the target information of identification, described telecommunication flow information acquisition method is carried out following steps:

Step 1 video image acquisition: by the video camera on the support that is arranged on road side or intermediate isolating band and the real-time video sequence image of image capture device collection;

Step 2 moving object detection: in the video sequence image obtained from step 1, motion target area is extracted from background image, adopt background subtraction to obtain the zone that preliminary target vehicle may exist, comprise the traffic digital image of target vehicle;

Step 3 traffic image pre-service: the traffic digital image obtained is carried out to the pre-service such as filtering, removal noise from step 2, eliminate noise and take threshold denoising;

Step 4 traffic image compression: adopt improved Embedded Zerotree Wavelet Coding to carry out compression of images to the processing traffic digital image later obtained from step 3 and obtain compressed bit stream;

The transmission of step 5 compressed bit stream and decompression: to the compressed bit stream process transmission incoming terminal obtained, and it is obtained to reconstructed image by decompression from step 4;

The telecommunication flow information data are added up and calculated to step 6 in the basic enterprising driving of step 5.

Wherein, the improved Embedded Zerotree Wavelet Coding of described employing comprises the following steps:

Step 41 adopts suitable wavelet basis to carry out wavelet transformation to traffic image;

Step 42 initialization threshold value

wherein f (i, j) is the wavelet conversion coefficient collection;

The scanning of step 43 main coding, form master meter;

The step 44 threshold value T=T/2 that reduces by half;

The coded scanning of step 45 subordinate, form auxiliary table;

Step 46 is sorted the order of the significant wavelet coefficients quantized value stored in auxiliary table according to the size of significant coefficient place interval value;

Step 47 forms new matrix of coefficients by the difference of coefficient of efficiency and reconstruction value and carries out the threshold value renewal;

Step 48 judgment threshold T=1, if set up execution step 49; The execution step 43 if be false;

Step 49 output encoder symbol stream.

Main coding scanning in described step 43 specifically comprises:

Current coefficient and threshold value are compared, if the absolute value of current coefficient is less than threshold value, and the threshold value of being greater than is arranged in its descendants's coefficient, output symbol Z, on the contrary, its descendants's coefficient all is less than threshold value, output symbol T; If the absolute value of current coefficient is greater than threshold value, then judge in descendants's coefficient of this coefficient whether the threshold value of being greater than is arranged, if output symbol P and N are arranged, otherwise output symbol P ₁and N ₁.In scanning process, these symbols are stored in a master meter, after the i time scanning, the locational coefficient of significant coefficient is set to 0, in upper these positions of directly skipping while once scanning.

Subordinate coded scanning in described step 45 specifically comprises:

The wavelet coefficient that is considered to significant coefficient in main coding scanning is carried out to quantization encoding.At first construct quantizer when quantizing, the interval of quantizer is [T _i-1, 2T _i-1), interval maximal value is current threshold value 2 times, minimum value is current threshold value; Input interval is divided into to two interval [T _i-1, 1.5T _i-1) and [1.5T _i-1, 2T _i-1), if coefficient belongs to quantized interval [T _i-1, 1.5T _i-1) be quantified as 0, if belong to interval [1.5T _i-1, 2T _i-1) be quantified as 1, and store these binit stream with an auxiliary table.

The accompanying drawing explanation

Fig. 1 video image acquisition process flow diagram.

Fig. 2 moving object detection process flow diagram.

Fig. 3 traffic image pretreatment process figure.

Fig. 4 traffic image compression process figure.

Fig. 5 Methods for Traffic Capacity at Signal Junction calculation flow chart.

Fig. 6 car statistics and traffic information data calculation flow chart.

Embodiment

Below in conjunction with accompanying drawing, be described further.

A kind of telecommunication flow information acquisition method based on compression of images, comprise that the video to gathering carries out moving object detection with background subtraction, digital traffic image is carried out to filtering and remove the pre-service such as noise, utilize improved Embedded Zerotree Wavelet Coding to carry out compression of images and the code stream after compression is carried out to data transmission and reconstructed image is obtained in decompression, carry out car statistics and calculate traffic information data according to the target information of identification, described telecommunication flow information acquisition method is carried out following steps:

Step 1 video image acquisition: by the video camera on the support that is arranged on road side or intermediate isolating band and the real-time video sequence image of image capture device collection;

Step 2 moving object detection: in the video sequence image obtained from step 1, motion target area is extracted from background image, adopt background subtraction to obtain the zone that preliminary target vehicle may exist, comprise the traffic digital image of target vehicle;

Step 3 traffic image pre-service: the traffic digital image obtained is carried out to the pre-service such as filtering, removal noise from step 2, eliminate noise and take threshold denoising;

Step 4 traffic image compression: adopt improved Embedded Zerotree Wavelet Coding to carry out compression of images to the processing traffic digital image later obtained from step 3 and obtain compressed bit stream;

The transmission of step 5 compressed bit stream and decompression: to the compressed bit stream process transmission incoming terminal obtained, and it is obtained to reconstructed image by decompression from step 4;

The telecommunication flow information data are added up and calculated to step 6 in the basic enterprising driving of step 5.

As shown in Figure 1, step 1 video image acquisition flow process comprises the following steps:

Step S101 observation station is chosen, and settles picture pick-up device;

Step S102 opens camera, obtains device parameter;

Step S103 arranges standard and the frame format of video;

Step S104 starts video acquisition;

Step S105 obtains the real-time traffic video.

As shown in Figure 2, the flow process of step 2 moving object detection comprises the following steps:

Step S201 carries out pre-service to the video image of input;

Step S202 background modeling also extracts background image;

Step S203 current video two field picture and Background subtract each other;

Step S204 binaryzation foreground picture;

Step S205 carries out medium filtering and morphologic filtering, removes noise;

Step S206 is converted into picture format by background, in order to show.

As shown in Figure 3, the pretreated flow process of step 3 traffic image comprises the following steps:

Step S301 carries out the filtering processing to traffic image;

Step S302 carries out threshold denoising to traffic image.

In described step S302, threshold value is definite, specifically comprises:

The white noise that is σ to noise intensity, soft-threshold T _nby following formula, determined:

$\mathrm{\&eta;}\left(\mathrm{\&omega;}\right)=\left\{\begin{array}{cc}\mathrm{\&omega;}+T_n-\frac{T_n}{2k+1},& \mathrm{\&omega;}<-T_n\\ \frac{1}{(2k+1)t^{2k}}{\mathrm{\&omega;}}^{2k+1},& \left|\mathrm{\&omega;}\right|\&le;T_n\\ \mathrm{\&omega;}-T_n+\frac{T_n}{2k+1},& \mathrm{\&omega;}>T_n\end{array}\right.$

Wherein, ω is original wavelet coefficients, and η (ω) means the wavelet coefficient after thresholding.

As shown in Figure 4, the flow process of step 4 traffic image compression comprises the following steps:

Step 401 adopts suitable wavelet basis to carry out wavelet transformation to traffic image;

Step 402 initialization threshold value

wherein f (i, j) is the wavelet conversion coefficient collection;

The scanning of step 403 main coding, form master meter;

The step 404 threshold value T=T/2 that reduces by half;

The coded scanning of step 405 subordinate, form auxiliary table;

Step 406 is sorted the order of the significant wavelet coefficients quantized value stored in auxiliary table according to the size of significant coefficient place interval value;

Step 407 forms new matrix of coefficients by the difference of coefficient of efficiency and reconstruction value and carries out the threshold value renewal;

Step 408 judgment threshold T=1, if set up execution step 409; The execution step 403 if be false;

Step 409 output encoder symbol stream.

Main coding scanning in described step 403 specifically comprises:

Current coefficient and threshold value are compared, if the absolute value of current coefficient is less than threshold value, and the threshold value of being greater than is arranged in its descendants's coefficient, output symbol Z, on the contrary, its descendants's coefficient all is less than threshold value, output symbol T; If the absolute value of current coefficient is greater than threshold value, then judge in descendants's coefficient of this coefficient whether the threshold value of being greater than is arranged, if output symbol P and N are arranged, otherwise output symbol P ₁and N ₁.In scanning process, these symbols are stored in a master meter, after the i time scanning, the locational coefficient of significant coefficient is set to 0, in upper these positions of directly skipping while once scanning.

Subordinate coded scanning in described step 405 specifically comprises:

The wavelet coefficient that is considered to significant coefficient in main coding scanning is carried out to quantization encoding.At first construct quantizer when quantizing, the interval of quantizer is [T _i-1, 2T _i-1), interval maximal value is current threshold value 2 times, minimum value is current threshold value; Input interval is divided into to two interval [T _i-1, 1.5T _i-1) and [1.5T _i-1, 2T _i-1), if coefficient belongs to quantized interval [T _i-1, 1.5T _i-1) be quantified as 0, if belong to interval [1.5T _i-1, 2T _i-1) be quantified as 1, and store these binit stream with an auxiliary table.

As shown in Figure 5, the flow process of the transmission of step 5 compressed bit stream and decompression comprises the following steps:

The transmission of step 501 compressed bit stream;

Step 502 receives compressed bit stream;

Step 503 arranges threshold value;

Step 504 structure inverse quantizer;

Step 505 is understood positional information and the wavelet coefficient information comprised in bit stream;

Step 506 wavelet inverse transformation;

Step 507 information loss tolerance, select suitable reconstructed image.

Information loss tolerance in described step 507 specifically comprises:

The method for objectively evaluating of employing based on reconstructed image and original image pixels difference criterion, can weigh with square error MSE and Y-PSNR PSNR:

$\mathrm{MSE}=\frac{1}{\mathrm{MN}}\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{[f(i,j)-f^{\&prime;}(i.j)]}^2$

$\mathrm{PSNR}=10\lg \frac{255^2}{\mathrm{MSE}}\left(\mathrm{dB}\right)$

Wherein, the size of original image is M * N, the gray shade scale of pixel be f (i, j) (i=1 ..., M; J=1 ..., N), the reconstructed image gray shade scale be f ' (i, j) (i=1 ..., M; J=1 ..., N); If PSNR>35dB, this reconstructed image meets the requirements.

As shown in Figure 6, step 6 car statistics the flow process that calculates traffic information data comprise the following steps:

Step S601 information of vehicles statistics;

Step S602 calculates the traffic flow datas such as the volume of traffic, speed, density.

Claims (7)

1. the telecommunication flow information acquisition method based on compression of images, comprise that the video to gathering carries out moving object detection with background subtraction, digital traffic image is carried out to filtering and remove the pre-service such as noise, utilize improved Embedded Zerotree Wavelet Coding to carry out compression of images and the code stream after compression is carried out to data transmission and reconstructed image is obtained in decompression, carry out car statistics and calculate traffic information data according to the target information of identification, described telecommunication flow information acquisition method is carried out following steps:

Step 1 video image acquisition: by the video camera on the support that is arranged on road side or intermediate isolating band and the real-time video sequence image of image capture device collection;

Step 2 moving object detection: in the video sequence image obtained from step 1, motion target area is extracted from background image, adopt background subtraction to obtain the zone that preliminary target vehicle may exist, comprise the traffic digital image of target vehicle;

Step 3 traffic image pre-service: the traffic digital image obtained is carried out to the pre-service such as filtering, removal noise from step 2, eliminate noise and take threshold denoising;

Step 4 traffic image compression: adopt improved Embedded Zerotree Wavelet Coding to carry out compression of images to the processing traffic digital image later obtained from step 3 and obtain compressed bit stream;

The transmission of step 5 compressed bit stream and decompression: to the compressed bit stream process transmission incoming terminal obtained, and it is obtained to reconstructed image by decompression from step 4;

The telecommunication flow information data are added up and calculated to step 6 in the basic enterprising driving of step 5.

Wherein, the improved Embedded Zerotree Wavelet Coding of described employing comprises the following steps:

Step 41 adopts suitable wavelet basis to carry out wavelet transformation to traffic image;

Step 42 initialization threshold value

wherein f (i, j) is the wavelet conversion coefficient collection;

The scanning of step 43 main coding, form master meter;

The step 44 threshold value T=T/2 that reduces by half;

The coded scanning of step 45 subordinate, form auxiliary table;

Step 46 is sorted the order of the significant wavelet coefficients quantized value stored in auxiliary table according to the size of significant coefficient place interval value;

Step 47 forms new matrix of coefficients by the difference of coefficient of efficiency and reconstruction value and carries out the threshold value renewal;

Step 48 judgment threshold T=1, if set up execution step 49; The execution step 43 if be false;

Step 49 output encoder symbol stream.

2. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, main coding scanning in step 43 specifically comprises:

Current coefficient and threshold value are compared, if the absolute value of current coefficient is less than threshold value, and the threshold value of being greater than is arranged in its descendants's coefficient, output symbol Z, on the contrary, its descendants's coefficient all is less than threshold value, output symbol T; If the absolute value of current coefficient is greater than threshold value, then judge in descendants's coefficient of this coefficient whether the threshold value of being greater than is arranged, if output symbol P and N are arranged, otherwise output symbol P ₁and N ₁.In scanning process, these symbols are stored in a master meter, after the i time scanning, the locational coefficient of significant coefficient is set to 0, in upper these positions of directly skipping while once scanning;

Subordinate coded scanning in step 45 specifically comprises:

The wavelet coefficient that is considered to significant coefficient in main coding scanning is carried out to quantization encoding.At first construct quantizer when quantizing, the interval of quantizer is [T _i-1, 2T _i-1), interval maximal value is current threshold value 2 times, minimum value is current threshold value; Input interval is divided into to two interval [T _i-1, 1.5T _i-1) and [1.5T _i-1, 2T _i-1), if coefficient belongs to quantized interval [T _i-1, 1.5T _i-1) be quantified as 0, if belong to interval [1.5T _i-1, 2T _i-1) be quantified as 1, and store these binit stream with an auxiliary table.

3. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, described step 1 video image acquisition flow process comprises the following steps:

Step S101 observation station is chosen, and settles picture pick-up device;

Step S102 opens camera, obtains device parameter;

Step S103 arranges standard and the frame format of video;

Step S104 starts video acquisition;

Step S105 obtains the real-time traffic video.

4. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, the flow process of step 2 moving object detection comprises the following steps:

Step S201 carries out pre-service to the video image of input;

Step S202 background modeling also extracts background image;

Step S203 current video two field picture and Background subtract each other;

Step S204 binaryzation foreground picture;

Step S205 carries out medium filtering and morphologic filtering, removes noise;

Step S206 is converted into picture format by background, in order to show.

5. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, the pretreated flow process of step 3 traffic image comprises the following steps:

Step S301 carries out the filtering processing to traffic image;

Step S302 carries out threshold denoising to traffic image;

In described step S302, threshold value is definite, specifically comprises:

The white noise that is σ to noise intensity, soft-threshold T _nby following formula, determined:

$\mathrm{\&eta;}\left(\mathrm{\&omega;}\right)=\left\{\begin{array}{cc}\mathrm{\&omega;}+T_n-\frac{T_n}{2k+1},& \mathrm{\&omega;}<-T_n\\ \frac{1}{(2k+1)t^{2k}}{\mathrm{\&omega;}}^{2k+1},& \left|\mathrm{\&omega;}\right|\&le;T_n\\ \mathrm{\&omega;}-T_n+\frac{T_n}{2k+1},& \mathrm{\&omega;}>T_n\end{array}\right.$

Wherein, ω is original wavelet coefficients, and η (ω) means the wavelet coefficient after thresholding.

6. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, the transmission of step 5 compressed bit stream and the flow process decompressed comprise the following steps:

The transmission of step 501 compressed bit stream;

Step 502 receives compressed bit stream;

Step 503 arranges threshold value;

Step 504 structure inverse quantizer;

Step 505 is understood positional information and the wavelet coefficient information comprised in bit stream;

Step 506 wavelet inverse transformation;

Step 507 information loss tolerance, select suitable reconstructed image;

Information loss tolerance in described step 507 specifically comprises:

The method for objectively evaluating of employing based on reconstructed image and original image pixels difference criterion, can weigh with square error MSE and Y-PSNR PSNR:

$\mathrm{MSE}=\frac{1}{\mathrm{MN}}\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{[f(i,j)-f^{\&prime;}(i.j)]}^2$

$\mathrm{PSNR}=10\lg \frac{255^2}{\mathrm{MSE}}\left(\mathrm{dB}\right)$

Wherein, the size of original image is M * N, the gray shade scale of pixel be f (i, j) (i=1 ..., M; J=1 ..., N), the reconstructed image gray shade scale be f ' (i, j) (i=1 ..., M; J=1 ..., N);

If PSNR>35dB, this reconstructed image meets the requirements.

7. a kind of telecommunication flow information acquisition method based on compression of images according to claim 1, is characterized in that, step 6 car statistics the flow process that calculates traffic information data comprise the following steps:

Step S601 information of vehicles statistics;

Step S602 calculates the traffic flow datas such as the volume of traffic, speed, density.

Images