CN105787886A - Multi-beam image sonar-based real-time image processing method - Google Patents

Abstract

The invention discloses a multi-beam image sonar-based real-time image processing method. According to the method, the multi-beam forward looking sonar image data of an underwater unmanned untethered submersible vehicle (AUV) are processed in real time, so that the real-time underwater obstacle information is provided for the AUV. In this way, the underwater obstacle-avoiding function of the AUV is realized. According to the technical scheme of the invention, based on the principle of the image sonar imaging technique, a large amount of noises and noise speckles in sonar images are filtered out through the dynamic distance attenuation threshold method filtering process and the enhancement algorithm. Meanwhile, the signal-to-noise ratio of images is improved. Based on the fuzzy clustering and morphology algorithm, the obstacle information in sonar images is filtered out from the sonar images. Therefore, the algorithm of the invention is stable, reliable and good in real-time performance. Based on the method, the obstacle information during the navigation process can be accurately and effectively provided. Furthermore, the underwater safety of the AUV is improved.

Description

A kind of real time image processing based on multi-beam image sonar

Technical field

The present invention relates to underwater robot technical field, particularly relate to a kind of real time image processing based on AUV forward sight multi-beam image sonar.

Background technology

Autonomous underwater robot (AutonomousUnderwaterVehicle, AUV) is a kind of unmanned, untethered, carry the energy, the intelligent robot of the task that independently fulfils assignment under water.Growing along with various application demands, requires also more and more higher to autonomous underwater robot working space: extend from known map, smooth bottom region to the unknown complex landform sea area of offshore, near Sea Bottom.Actual circumstances not known introduces a large amount of uncertain to autonomous underwater robot, and wherein that it is dangerous maximum is unknown obstacle: the ridge of those seabed projections, coral reef, shipwreck, submarine mine, mooring platform, artificial structure's thing and submarine, large ocean biology are all likely to bring fatal harm to autonomous underwater robot.How fast and accurately the unknown obstacle in the actual marine environment of perception is one of restriction wide variety of key issue of autonomous underwater robot.AUV to realize hiding unknown obstacle in real time, it is necessary to possesses real-time perception and obtains the ability of circumstances not known complaint message.Therefore, the scan picture of forward sight image sonar obtains the ability of complaint message, is that AUV completes Realtime collision free decision-making and completes basis and the premise of mission task.

Summary of the invention

For real-time perception and the ability obtaining circumstances not known complaint message, the technical problem to be solved in the present invention is to utilize AUV forward sight anticollision sonar Real-time Collection sonar image, and by real time image processing extracting underwater obstacle target information.

The present invention is the technical scheme is that a kind of real time image processing based on multi-beam image sonar for achieving the above object, comprises the following steps:

Carry out image filtering process, filter the noise in image and noise speckle；

Carry out image enhancement processing, improve signal noise ratio (snr) of image；

Carry out image blurring clustering processing, the information of barrier is split from sonar image；

Carry out morphological image process, remove less speck, the leak filling up image and crack.

Described image filtering is processed the software tool kit being to be carried by image sonar and obtains real-time sonar image, and sonar image is projected in latent device coordinate system, thus obtaining the gray value of each pixel in image and the distance of each pixel distance sonar, and according to the relation between sonar operating range and target-echo intensity, adopt dynamic distance attenuation threshold method that image is filtered.

Described image enhancement processing is by processing filtered sonar image, obtain maximin in image, dynamic distance attenuation threshold method is adopted with the attenuation relation of distance according to obstacle echo strength, gray value pixel more than threshold value is made to be enhanced, to improve signal to noise ratio, thus the profile information of underwater obstruction in prominent image.

Described dynamic distance attenuation threshold method, is added in thresholding with pad value 20 lgR of distance by sound wave, makes the thresholding that this filtering is not single filter, but threshold value increases with propagation distance and decays, wherein R is sonic propagation distance.

Described fuzzy clustering processing method adopts K means clustering algorithm, n data object is divided into k class, makes to gather in class, evacuates between class, adopts mean square deviation as similarity measure function, comprises the following steps:

1) determine clusters number k, arbitrarily select k object as initial cluster center；

2) for each sample, finding the cluster centre that it is nearest, and assign it to closest apoplexy due to endogenous wind, cluster centre is expressed as:

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j\∈\{\mathrm{1,2}...k\}}{\min }{\left|\right|x_i-p_j\left|\right|}^2$

Wherein k is cluster centre number, and i is the ith pixel point in sonar image, and j is the jth cluster centre in K, x_iFor the gray value of ith pixel point, p in image_jNumerical value for jth cluster centre；

3) recalculating the new cluster centre updating every class, continuous iteration is till target function type reaches minima and cluster centre no longer changes.

Described Morphological scale-space method comprises the following steps:

1) corrosion treatmentCorrosion Science, is used for filtering noise:

$\mathrm{I\ΘJ}=\left\{x\right|J_x\⋐I\}$

Wherein, I is image sonar original image, and J is structural element, and x is any point being not zero in original bianry image, and structural element J obtains J after translating x_xIf, J_xBe contained in I, then the point set of all x point compositions meeting above-mentioned condition is called the I result corroded by J；

2) expansion process, for the profile of smoothed image, the gap filled up on profile, particularly as follows:

$I\⊕J={\left[I^C\mathrm{\Θ}(-J)\right]}^C$

That is, relative to initial point, J being rotated 180 ° and obtain-J ,-J is to I for recycling^CCorroding, the supplementary set of Corrosion results is exactly expansion results；Wherein, I is image sonar original image, and J is structural element, I^CRepresent the supplementary set of I.

The present invention has the following advantages and beneficial effect:

1. the present invention is based on adopting the filtering of dynamic distance attenuation threshold method on image sonar image-forming principle basis and strengthening algorithm by much noise in sonar image and noise speckle elimination, and improves signal noise ratio (snr) of image.

2. utilize fuzzy clustering and Morphology Algorithm the information of barrier in sonar image to be split from sonar image.

3. the algorithmic stability being previously mentioned in the present invention is reliable, real-time good, can provide AUV complaint message in navigation process accurately and effectively, and then improve the underwater safety of AUV.

4. adopt and first corrode the computing (i.e. opening operation on mathematical morphology) expanded afterwards, because what be most difficult to during sonar image to remove is exactly the relatively very big speckle of echo strength, so process and can remove less bright spot, the burrs on edges removing image and isolated point, the leak filling up image and crack, retain all of gray scale and bigger clear zone feature invariant simultaneously.

5. image filtering is processed the software tool kit being to be carried by image sonar and obtains real-time sonar image, obtained gray value and the distance value of each pixel in image by algorithm, and dynamic threshold is set according to the relation between sonar operating range and target-echo intensity has arrived the function of image filtering.This processing method can remove the noise that major part is done by transducer sensitivity difference, temperature fluctuation, marine environment, many ways, quantified the forms such as Gaussian noise that the factor such as uneven formed, salt-pepper noise, speckle noise.

Accompanying drawing explanation

Fig. 1 is the composition schematic diagram of the present invention；

Fig. 2 is the image processing flow figure of the present invention；

Fig. 3 is the algorithm flow chart of the present invention.

Detailed description of the invention

Below in conjunction with accompanying drawing and example, the present invention is described in further detail.

The present invention is made up of the multi-beam image sonar (for Real-time Collection sonar image data) being equipped on latent device and pattern process computer (real-time acquisition processes multi-beam image sonar data), as it is shown in figure 1, include:

Multi-beam image sonar, is equipped on latent device, for Real-time Collection sonar image data；

Device sonar image of diving processes computer, connect multi-beam image sonar, for obtaining record multi-beam image sonar data in real time, and obtain real-time complaint message in Forward-looking Sonar angular field of view by scan picture software, and complaint message is sent top level control computer.

Pattern process computer is by connecting multi-beam image sonar, for obtaining in real time, processing multi-beam image sonar data, the software tool kit that image processing software is carried by image sonar obtains real-time sonar image, carry out image filtering by algorithm and enhancing algorithm by much noise in sonar image and noise speckle elimination and improves signal noise ratio (snr) of image, and utilize fuzzy clustering and Morphology Algorithm the information of barrier in sonar image to be split from sonar image, as shown in Figure 2.Wherein said fuzzy clustering processing method particularly as follows:

N data object is reasonably divided into k class by K means clustering algorithm, makes to gather in class, evacuates between class.Adopt mean square deviation as similarity measure function.Algorithm flow is as follows:

1) determine clusters number k, arbitrarily select k object as initial cluster center.

2) for each sample, finding the cluster centre that it is nearest, and assign it to closest apoplexy due to endogenous wind, cluster centre is expressed as:

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j\∈\{\mathrm{1,2}...k\}}{\min }{\left|\right|x_i-p_j\left|\right|}^2$

Wherein k is cluster centre number, is the i-th sample point, for jth cluster centre value.

3) recalculating the new cluster centre updating every class, continuous iteration is till target function type reaches minima and cluster centre no longer changes, and algorithm flow chart is as shown in Figure 3.

Adopt and first corrode the computing (i.e. opening operation on mathematical morphology) expanded afterwards, because what be most difficult to during sonar image to remove is exactly the relatively very big speckle of echo strength, so process and can remove less bright spot, the burrs on edges removing image and isolated point, the leak filling up image and crack, retain all of gray scale and bigger clear zone feature invariant simultaneously.

Claims (6)

1. the real time image processing based on multi-beam image sonar, it is characterised in that comprise the following steps:

Carry out image filtering process, filter the noise in image and noise speckle；

Carry out image enhancement processing, improve signal noise ratio (snr) of image；

Carry out image blurring clustering processing, the information of barrier is split from sonar image；

Carry out morphological image process, remove less speck, the leak filling up image and crack.

2. a kind of real time image processing based on multi-beam image sonar according to claim 1, it is characterized in that, described image filtering is processed the software tool kit being to be carried by image sonar and obtains real-time sonar image, and sonar image is projected in latent device coordinate system, thus obtaining the gray value of each pixel in image and the distance of each pixel distance sonar, and according to the relation between sonar operating range and target-echo intensity, adopt dynamic distance attenuation threshold method that image is filtered.

3. a kind of real time image processing based on multi-beam image sonar according to claim 1, it is characterized in that, described image enhancement processing is by processing filtered sonar image, obtain maximin in image, dynamic distance attenuation threshold method is adopted with the attenuation relation of distance according to obstacle echo strength, gray value pixel more than threshold value is made to be enhanced, to improve signal to noise ratio, thus the profile information of underwater obstruction in prominent image.

4. a kind of real time image processing based on multi-beam image sonar according to Claims 2 or 3, it is characterized in that, described dynamic distance attenuation threshold method, it is that sound wave is added in thresholding with pad value 20 lgR of distance, the thresholding that this filtering is not single is made to filter, but threshold value increases with propagation distance and decays, wherein R is sonic propagation distance.

5. a kind of real time image processing based on multi-beam image sonar according to claim 1, it is characterized in that, described fuzzy clustering processing method adopts K means clustering algorithm, n data object is divided into k class, make to gather in class, evacuate between class, adopt mean square deviation as similarity measure function, comprise the following steps:

1) determine clusters number k, arbitrarily select k object as initial cluster center；

2) for each sample, finding the cluster centre that it is nearest, and assign it to closest apoplexy due to endogenous wind, cluster centre is expressed as:

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j\∈\{\mathrm{1,2}...k\}}{\min }{\left|\right|x_i-p_j\left|\right|}^2$

Wherein k is cluster centre number, and i is the ith pixel point in sonar image, and j is the jth cluster centre in K, x_iFor the gray value of ith pixel point, p in image_jNumerical value for jth cluster centre；

3) recalculating the new cluster centre updating every class, continuous iteration is till target function type reaches minima and cluster centre no longer changes.

6. a kind of real time image processing based on multi-beam image sonar according to claim 1, it is characterised in that described Morphological scale-space method comprises the following steps:

1) corrosion treatmentCorrosion Science, is used for filtering noise:

$\mathrm{I\ΘJ}=\left\{x\right|J_x\⋐I\}$

Wherein, I is image sonar original image, and J is structural element, and x is any point being not zero in original bianry image, and structural element J obtains J after translating x_xIf, J_xBe contained in I, then the point set of all x point compositions meeting above-mentioned condition is called the I result corroded by J；

2) expansion process, for the profile of smoothed image, the gap filled up on profile, particularly as follows:

$\mathrm{I\ΘJ}={\left[I^C\mathrm{\Θ}(-J)\right]}^C$

That is, relative to initial point, J being rotated 180 ° and obtain-J ,-J is to I for recycling^CCorroding, the supplementary set of Corrosion results is exactly expansion results；Wherein, I is image sonar original image, and J is structural element, I^CRepresent the supplementary set of I.