CN106296603A - A kind of small target detecting method of underwater sonar image - Google Patents

Abstract

The invention belongs to object detection field, be specifically related to the small target detecting method of a kind of underwater sonar image.The present invention includes: initialize: making binary-state threshold th=0.8, shade direction is θ degree, makes I₀=I；Make SE₁Represent To Template, SE₁It is the square templates of a size of 3 × 3 pixels, makes SE₀Represent background template, SE₀It it is the marginal area of the square templates of a size of 25 × 25 pixels.Target detection template proposed by the invention is simple；Detection speed is fast；Noise resisting ability is strong；Without training data.

Description

A kind of small target detecting method of underwater sonar image

Technical field

The invention belongs to object detection field, be specifically related to the small target detecting method of a kind of underwater sonar image.

Background technology

Underwater sonar system, except the competent extreme operating environments of needs, should also have efficient Target detection and identification energy Power, it is possible to accurately differentiate and sink to the bottom Small object and water Small Target.But, owing to sonar system is at the echo of certain reception, it is Respectively reflecting the vector of sound wave, the echo amplitude of system record is wobble variation, produces speckle accordingly when forming sound spectrogram Noise, the impact explanation to sound spectrogram.Although along with imaging sonar technology reaches its maturity, side-scan sonar can provide high resolution graphics Picture, but compared with optical imagery, side-scanning sonar image still has that image quality is poor, between target and background, contrast is low etc. Problem.Especially for small target deteection problem under water, due to interchannel noise and the impact of shot noise so that existing target Detection technique is extremely difficult to satisfied result.

Image processing techniques and mode identification technology is utilized to detect Small object from underwater sonar data from 90 years last century Achieved with rapid progress since generation.These methods can be generally divided into two classes: has measure of supervision and unsupervised approaches.Wherein have Measure of supervision needs a large amount of training datas, and algorithm performance height relies on the character of training data.This kind of method mainly includes that K-is near Adjacent neutral net, optimal identification filtered classification device, Bayes classifier, principal component analysis etc..On the other hand, unsupervised approaches without Need special training data, it is adaptable to the input data of comparatively wide scope, data-optimized not for special training.This type of method Including Markov random field, the statistics technology such as snake model, Morphological scale-space.Additionally, in order to improve algorithm performance further, also Having scholar various different detection/recognition technology to be blended, the main integration technology used includes: by the inspection of various algorithms The logic that uses survey/recognition result merges (such as boolean and or computing), log-likelihood ratio testing algorithm, linear weighted function fusion etc..

The present invention utilizes the morphological transformation occupation rate with stronger anti-noise ability to hit-miss conversion (Percentage Occupancy Hit-or-Miss Transform, POHMT), it is achieved sound is swept in one side fast and accurately Receive image small target deteection algorithm.

Summary of the invention

It is an object of the invention to provide one and hit-miss conversion (Percentage based on occupation rate Occupancy Hit-or-Miss Transform, POHMT) the small target detecting method of underwater sonar image.

The object of the present invention is achieved like this:

(1) initialize: making binary-state threshold th=0.8, shade direction is θ degree, makes I₀=I, wherein I is H × W pixel Input sonar image matrix；

(2) SE is made₁Represent To Template, SE₁It is the square templates of a size of 3 × 3 pixels, makes SE₀Represent background mould Plate, SE₀It it is the marginal area of the square templates of a size of 25 × 25 pixels；

(3) image I is calculated₀Luminance mean value, use m₀Represent, make t=m₀... 255, for each t, calculate bianry image B_t:

Wherein, h=1 ... H, w=1 ... W；

(4) foreground template matching rate poFore and background template matching rate poBack is calculated respectively

PoFore (t)=conv (B_t,SE₁)/9

PoBack (t)=conv (1-B_t,SE₀)/96

Wherein, t=m₀... 255, conv (.) represents convolution algorithm, for fixing t value, poFore (t) and poBack T () is H × W matrix；

(5) image I is calculated₀Occupation rate hit-miss conversion poHMT

$poHMT=\underset{t=m\mathrm{0....255}}{max}\left\{\min \right\{poFore\left(t\right),poBack\left(t\right)\left\}\right\}$

Wherein, min{.} is minimum value function, min{poFore (t), poBack (t) } represent for a fixing t Value, takes less among both front poFore (t) and poBack (t) one；Max{.} is max function；PoHMT is H × W Matrix；

(6) poHMT of step (5) is done binary conversion treatment

Wherein, h=1 ... H, w=1 ... (h w) represents that two-dimensional matrix poHMT is at (h, w) value at place, poBW for W, poHMT For two values matrix, (h w) represents that poBW is in (h, w) value at place to poBW；

(7) morphological dilations, erosion operation is utilized to remove the tiny noise spot in poBW；

(8) poBW is made₁Represent the poBW after denoising, utilize 8-join domain labeling method labelling poBW₁In connected region Territory, records the position of the boundary rectangle of each connected region, length and width respectively；

(9) I is made₀=255-I, performs step (3)～step (7), and uses poBW₀Represent the poBW after denoising, utilize 8- Join domain labeling method labelling labelling poBW₀In connected region, record respectively the boundary rectangle frame of each connection position, Length and width；

(10) with poBW₁In connected region R₁Boundary rectangle be foundation, determine region of search,；This is searched for Area maps is to poBW₀Respective regions, and represent poBW with RS₀In this region of search；

(11) at poBW₀RS region in scan for, if it is possible to find connected region R₀, then by poBW₁In Connected region R₁It is labeled as target, deletes region R simultaneously₀；Otherwise, if at poBW₀RS in there is not connected region, then will Connected region R in poBW1₁It is labeled as non-targeted.

(12) repeat step (10)～step (11), check poBW₁In next connected region, until traversal poBW₁'s All connected regions.

The beneficial effects of the present invention is: 1) the target detection template that proposed is simple；2) detection speed is fast；3) antinoise Ability is strong；4) without training data.

Accompanying drawing explanation

Fig. 1 is the target in side-scanning sonar image and shade schematic diagram thereof；

Fig. 2 is To Template schematic diagram；

Fig. 3 is background template schematic diagram；

Fig. 4 is region of search schematic diagram.

Detailed description of the invention

Below in conjunction with the accompanying drawings the present invention is described further.

The present invention proposes one and the technical scheme is that 1) initialize: detection threshold value is set, determines target and background template Size, shade direction is set；2) former sonar image is carried out POHMT computing, and operation result is done binary conversion treatment, obtain mesh Mark bianry image；3) former sonar image does inverse process；4) sonar image after inverse is done POHMT computing, and will fortune Calculate result and do binary conversion treatment, obtain shade bianry image；5) target bianry image, shade bianry image and shade side are utilized To, mark Small object.

If I is input sonar image matrix, a size of H × W pixel, I (h, w) represent (h, w) brightness of position image, Wherein h=1,2 ..., H, w=1,2 ..., W；And comprising multiple high luminance targets region in image I, each target is respectively provided with Corresponding low-light level shadow region, as shown in Figure 1.

Object detection method proposed by the invention to implement step as follows:

Step 1, initializes: making binary-state threshold th=0.8, shade direction is θ degree, makes I₀=I.

Step 2, makes SE₁Represent To Template, here SE₁It is the square templates of a size of 3 × 3 pixels, as shown in Figure 2； Make SE₀Represent background template, here SE₀It is the marginal area of the square templates of a size of 25 × 25 pixels, as it is shown on figure 3, figure In 3, the size of gray area is 23 × 23 pixels, and its pixel value is 0, and the pixel value of edge white portion is 1.

Step 3, calculates image I₀Luminance mean value, use m₀Represent；Make t=m₀... 255, for each t, according to formula (1) bianry image B is calculated_t:

Wherein, h=1 ... H, w=1 ... W.

Step 4, utilizes formula (2) to calculate foreground template matching rate poFore, utilizes formula (3) to calculate background template coupling Rate poBack:

PoFore (t)=conv (B_t,SE₁)/9 (2)

PoBack (t)=conv (1-B_t,SE₀)/96 (3)

Wherein, t=m₀... 255, conv (.) represents convolution algorithm, for fixing t value, poFore (t) and poBack T () is H × W matrix.

Step 5, utilize formula (4) calculate image occupation rate hit-miss conversion poHMT:

$poHMT=\underset{t=m\mathrm{0....255}}{max}\left\{\min \right\{poFore\left(t\right),poBack\left(t\right)\left\}\right\}---\left(4\right)$

Wherein, min{.} is minimum value function, min{poFore (t), poBack (t) } represent for a fixing t Value, takes less among both poFore (t) and poBack (t) one；Max{.} is max function；PoHMT is H × W square Battle array.

Step 6, utilizes formula (5) that the poHMT of step 5 is done binary conversion treatment

Wherein, h=1 ... H, w=1 ... (h w) represents that two-dimensional matrix poHMT is at (h, w) value at place, poBW for W, poHMT For two values matrix, poBW (h, w) represent poBW (h, w) value at place,

Step 7, utilizes morphological dilations, erosion operation to remove the tiny noise spot in poBW.

Step 8, makes poBW₁Represent the poBW after denoising, utilize 8-join domain labeling method labelling poBW₁In company Logical region, records the position of the boundary rectangle of each connected region, length and width respectively.

Step 9, makes I₀=255-I, performs step 3～step 7, and uses poBW₀Represent the poBW after denoising, utilize 8- Join domain labeling method labelling labelling poBW₀In connected region, record the position of the boundary rectangle frame of each connected region respectively Put, length and width.

Step 10, with poBW₁In connected region R₁Boundary rectangle be foundation, determine region of search, such as Fig. 4 institute Show, wherein w=5, l=5 × w₀, w₀For connected region R₁Boundary rectangle width.This region of search is mapped to poBW₀Phase Answer region, and represent poBW with RS₀In this region of search.

Step 11, at poBW₀RS region in scan for, if it is possible to find a connected region (to assume to use R₀Represent This connected region), then by poBW₁In connected region R₁It is labeled as target, deletes region R simultaneously₀；Otherwise, if poBW₀RS in there is not connected region, then by connected region R in poBW1₁It is labeled as non-targeted.

Step 12, repeats step 10～step 11, checks poBW₁In next connected region, until traversal poBW₁'s All connected regions.

Claims (1)

1. the small target detecting method of a underwater sonar image, it is characterised in that comprise the steps:

(1) initialize: making binary-state threshold th=0.8, shade direction is θ degree, makes I₀=I, wherein I is the input of H × W pixel Sonar image matrix；

(2) SE is made₁Represent To Template, SE₁It it is the square templates of a size of 3 × 3 pixels；Make SE₀Represent background template, SE₀ It it is the marginal area of the square templates of a size of 25 × 25 pixels；

(3) image I is calculated₀Luminance mean value, use m₀Represent；Make t=m₀... 255, for each t, calculate bianry image B_t:

Wherein, h=1 ... H, w=1 ... W；

(4) foreground template matching rate poFore and background template matching rate poBack is calculated respectively:

PoFore (t)=conv (B_t,SE₁)/9

PoBack (t)=conv (1-B_t,SE₀)/96

Wherein, t=m₀... 255, conv (.) represents convolution algorithm；

(5) image I is calculated₀Occupation rate hit-miss conversion poHMT:

$poHMT=\underset{t=m\mathrm{0....255}}{max}\left\{\min \right\{poFore\left(t\right),poBack\left(t\right)\left\}\right\}$

Wherein, min{.} is minimum value function, and max{.} is max function；

(6) poHMT of step (5) is done binary conversion treatment:

Wherein, h=1 ... H, w=1 ... W, poHMT (h, w) represent two-dimensional matrix poHMT (h, w) value at place, poBW (h, w) Represent that poBW is in (h, w) value at place；

(7) morphological dilations, erosion operation is utilized to remove the tiny noise spot in poBW；

(8) poBW is made₁Represent the poBW after denoising, utilize 8-join domain labeling method labelling poBW₁In connected region, Record the position of the boundary rectangle of each connected region, length and width respectively；

(9) I is made₀=255-I, performs step (3)～step (7), and uses poBW₀Represent the poBW after denoising, utilize 8-to connect Zone marker method labelling poBW₀In connected region, record the position of the boundary rectangle of each connected region, length and width respectively；

(10) with poBW₁In connected region R₁Boundary rectangle be foundation, determine region of search,；By this region of search It is mapped to poBW₀Respective regions, and represent poBW with RS₀In this region of search；

(11) at poBW₀RS region in scan for, if it is possible to find connected region R₀, then by poBW₁In connection Region R₁It is labeled as target, deletes region R simultaneously₀；Otherwise, if at poBW₀RS in there is not connected region, then by poBW₁ In connected region R₁It is labeled as non-targeted.

(12) repeat step (10)～step (11), check poBW₁In next connected region, until traversal poBW₁All Connected region.