CN103632373A - Floc detection method combining three-frame differential higher-order statistics (HOS) with OTSU algorithm - Google Patents
Floc detection method combining three-frame differential higher-order statistics (HOS) with OTSU algorithm Download PDFInfo
- Publication number
- CN103632373A CN103632373A CN201310658750.3A CN201310658750A CN103632373A CN 103632373 A CN103632373 A CN 103632373A CN 201310658750 A CN201310658750 A CN 201310658750A CN 103632373 A CN103632373 A CN 103632373A
- Authority
- CN
- China
- Prior art keywords
- floc
- image
- target
- frame
- flco
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
Disclosed is a floc detection method combining three-frame differential higher-order statistics (HOS) with an OTSU algorithm. Since a commonly used floc tracking method is susceptible to noise, light and floc movement speed and the like at present and complete characteristics of a floc movement target are hard to extract, a floc target detection method based on combining the three-frame differential HOS with a particle swarm optimization enhancing OTSU algorithm is put forward. The floc detection method includes: subjecting consecutive three-frame images to differential calculation; calculating fourth-order moment pixel by pixel and comparing the fourth-order moment with threshold values; utilizing particle swarm optimization enhancing OTSU algorithm to acquire a best threshold value; utilizing the best threshold value to perform image binaryzation and image post-processing to acquire a relatively clear floc target finally. Therefore, a foundation of automatic subsequent floc analysis is laid. The floc detection method has the advantages of being accurate and rapid, capable of effectively extracting the floc target and applicable to effective extraction of the floc target in water treatment.
Description
Technical field
The present invention relates to a kind of three-frame difference high-order statistic in conjunction with the flco detection method of OTSU algorithm, belong to water treatment flco detection method technical field.
Background technology
Along with improving constantly of people's living standard, people are also more and more higher to the requirement of drinking water quality, and conventional water treatment process generally comprises several stages such as coagulation, precipitation, filtration, sterilization both at home and abroad.In the process of water treatment, can produce the flco phenomenon of condensing, and flco number, size, settling velocity etc. are the important parameters that judges coagulation effect.Moving object detection is the essential step in floc image quantitative test, be also affect that flco is analysis automated, the key of job stability, result accuracy.The situations such as uneven illumination is even owing to may existing in imaging process for image that industrial camera obtains, flco movement velocity slow, flco changes in distribution, the shade of flco is differed, lack of homogeneity, have noise, and also there is the different phenomenon of surface reflection in flco simultaneously.Therefore realizing exactly the detection of flco target and background, is the basis of further analyzing floc characteristic.
Realizing at present flco order object detection method has a lot, the conventional method of difference of having powerful connections, frame differential method and optical flow method.
Frame differential method is between in video sequence 2 or 3 consecutive frames, adopts and extracts the moving region in image by pixel difference thresholding, and it can detect the part changing in video image faster.Frame difference method method is simple, and relatively good to the adaptability of environment, stability is high, in the ideal case, when video camera is static, if subtract each other gray scale difference between rear continuous hardwood, is zero, thinks that this point belongs to static background, otherwise belongs to moving object region.Whether but in actual conditions, often having a lot of noise, is only zero situation of change that can not determine this point according to difference.Three-frame difference adopts three two field pictures to subtract each other respectively, and effect is better than two frame differences.Though simply can substantial separation go out moving target and background to error image thresholding method in frame difference method, the establishing method of threshold value is often comparatively difficult, and the impact of very difficult filtering noise.So need further to improve to the disposal route of noise and improve the choosing method of threshold value is carried out to effective disengaging movement target and background.
For method for processing noise, current have a lot of disposal routes, as smoothing denoising etc., wherein adopt high-order statistic (higher order statistics, HOS), refer to the statistic of 3 rank or the 3 above exponent numbers in rank, comprise Higher Order Cumulants, High Order Moment and higher-order spectrum, Higher Order Cumulants can suppress the impact of Gaussian noise and other some characteristics in theory completely, therefore, conventionally utilizes more Higher Order Cumulants and higher-order spectrum as processing noise instrument.
For Research on threshold selection, mainly contain bimodal method, process of iteration, maximum variance between clusters (OTSU, i.e. large Tianjin method).Although bimodal method is simple, adaptability is poor; Process of iteration operand is huge, is not suitable for real-time system; The effect of large Tianjin method is best comparatively speaking.First large Tianjin method is divided into two classes image pixel with threshold value T, then calculates inter-class variance and the class internal variance of two class grey scale pixel values, take ratio between two definite threshold T during as maximum.
But when simulated experiment, find, when processing background and be more or less the same with foreground target gray scale, by large Tianjin method, there will be a large amount of black regions, in the time of seriously, even can lose whole target.There is scholar to propose to introduce the large Tianjin of enhancing method that a kind of gray scale stretches, adopt nonlinear transformation to carry out gray scale stretching to image, definite threshold then, but this has also increased operation time.
Particle swarm optimization algorithm is a kind of probabilistic search method of finding in the process that group behavior is simulated.Algorithm is from primary group, by the optimum solution of iterative search particle fitness function.In iteration each time, the optimum solution gbest that the optimum solution pbest that each particle finds according to self and whole population find adjusts speed and the direction of motion, with the position of new particle more.
It is exactly to utilize particle swarm optimization algorithm to be optimized strengthening large Tianjin method that particle group optimizing strengthens large Tianjin method (OTSU), it has effectively solved the matter of time that strengthens large Tianjin method, first algorithm gives a RANDOM SOLUTION to the position of particle and speed, owing to being that gray level image scope is between 0-255, so will be within the scope of this given particle position, all random values are produced by rand () function.Then calculate fitness function value and compare, iterative computation, finally tries to achieve optimum solution, namely optimal threshold.
Summary of the invention
The present invention seeks to, the problem existing for the deficiency of three-frame difference method and noise effect and current Research on threshold selection, the present invention proposes and a kind ofly based on three three-frame differences, divide high-order statistic (HOS) and particle group optimizing to strengthen the flco object detection method of large Tianjin method (OTSU).
Realizing technical scheme of the present invention is, the present invention is based on motion flco irregular, and large reason affected by environment is used based on three-frame difference, high-order statistic and particle group optimizing, to strengthen large Tianjin method and extract floc characteristic, comprises the following steps:
(1) respectively to present frame I
t, former frame I
t-1, a rear frame I
t+1carry out pre-service, comprise gray scale correction, the processing such as figure image intensifying;
(2) I
t-1, I
t, I
t+1do respectively difference, obtain difference image D
t+1, D
t;
(3) difference image is carried out obtaining difference image D with computing
i;
(4) utilize particle group optimizing to strengthen large Tianjin method and determine optimal threshold T;
(5) by pixel calculate Fourth-order moment and with optimal threshold T comparison, be greater than T value, think that it is moving target, pixel value is set to 1, otherwise is 0;
(6) to the imagery exploitation Image Post-processing Techniques in step (5), further improve incomplete border, thereby obtain complete boundary image, thereby realize the smooth extraction of flco target.
The method of described definite optimal threshold, first gives a RANDOM SOLUTION to the position of particle and speed, due to be gray level image scope between 0-255, so will be within the scope of this given particle position, all random values are produced by rand () function.Then calculate fitness function value and compare, iterative computation, finally tries to achieve optimum solution, namely optimal threshold.
Described Image Post-processing Techniques, first does once " opening " computing (Open Operation) with less structural element to image and eliminates isolated noise; With larger structural element, image is done once to " closing " computing (Close Operation) again and fill up target internal cavity; Further improve incomplete border, thereby obtain complete boundary image, thereby realize the smooth extraction of flco target.
The present invention is based on flco motion characteristics, proposed to strengthen based on three-frame difference high-order statistic (HOS) and particle group optimizing the method that large Tianjin method is carried out flco detection and extracted.Compare with existing scheme, the improved method of the present invention can be extracted the edge of motion flco more exactly, the flco target area of extracting is also more complete, especially good to the larger target extraction effect of area, solved to a certain extent the deficiency of traditional two frame difference methods, imperfect with noise in testing process, rim detection, easily produce the shortcoming in target " cavity ", can accurately realize the extract real-time of flco target area.
The invention has the beneficial effects as follows, the present invention adopts three-frame difference method to extract flco target, has avoided the target that two frame difference methods may occur to extract the problems such as imperfect; But in three-frame difference method, there is On The Choice and the noise effect problem of optimal threshold simultaneously, the present invention adopts population to strengthen large Tianjin method and chooses optimal threshold, efficiently solving the threshold value that methods such as adopting process of iteration may occur chooses unreasonable, the problems such as efficiency is low, and adopt HOS technology, avoided the interference of noise.Therefore, the advantage such as improving one's methods that the present invention proposes has fast, accurately, extracts flco target complete, and noiseproof feature is good, this invention has certain practical value, can effectively extract flco target.
The present invention is applicable to the target of motion flco in water treatment and extracts.
Accompanying drawing explanation
Fig. 1 is preferably experimental facilities schematic diagram of the present invention;
Fig. 2 is process flow diagram of the present invention;
Fig. 3 is the contrast effect figure of method emulation in the present invention:
Fig. 3 (a) is original image;
Fig. 3 (b) is original image;
Fig. 3 (c) is original image;
Fig. 3 (d) is for to divide two three-frame differences of original image Fig. 3 (a);
Fig. 3 (e) is for to divide two three-frame differences of original image Fig. 3 (b);
Fig. 3 (f) is for to divide two three-frame differences of original image Fig. 3 (c);
Fig. 3 (g) is the divide+iteration of three three-frame differences to original image Fig. 3 (a);
Fig. 3 (h) is the divide+iteration of three three-frame differences to original image Fig. 3 (b);
Fig. 3 is (i) the divide+iteration of three three-frame differences to original image Fig. 3 (c);
Fig. 3 (j) is three three-frame differences of original image Fig. 3 (a) are divided+large Tianjin method;
Fig. 3 (k) is three three-frame differences of original image Fig. 3 (b) are divided+large Tianjin method;
Fig. 3 (l) is three three-frame differences of original image Fig. 3 (c) are divided+large Tianjin method;
Fig. 3 (m) is the disposal route of the present invention to original image Fig. 3 (a);
Fig. 3 (n) is the disposal route of the present invention to original image Fig. 3 (b);
Fig. 3 (o) is the disposal route of the present invention to original image Fig. 3 (c).
Embodiment
According to the step of the inventive method, the specific embodiment of the invention is as follows:
The first step: shown in Fig. 1, put experiment equipment.Sensor installation in flocculation basin end water; Make current level, the sampling window of flowing through lentamente, by current (flco) image of industrial camera continuous acquisition sampling window, in experiment, set according to the actual requirements sampling time interval.The time interval Tsampe adopting in this experiment be 1s(at 1s in the time, the flco number of observation forms has good representativeness) population is 100 left and right, can be used as sample for the treatment of image.
Second step: three-frame difference and noise processed.Former frame f to the image gathering respectively
t-1(x, y), present frame f
t(x, y), a rear frame f
t+1(x, y) carries out smoothing denoising, then calculates f
t-1(x, y) and f
t(x, y) difference image D
t, f
t+1(x, y) and f
t(x, y) difference image D
t+1, then by D
t, D
t+1carry out with computing after obtain difference image Q
diff(x, y), finally chooses appropriate threshold, for three-frame difference method, chooses rational threshold value the effect of target extraction is played an important role, and according to formula (6), difference result is carried out to thresholding and obtains two-value result.；
(A) suppose that extraneous illumination condition does not change or changes under little condition, being expressed as follows of adjacent image sequence:
f
t(x,y)＝M
t(x,y)+B
t(x,y)+n
t(x,y) (1)
(B) two formulas obtain the difference image of adjacent two two field pictures and are after subtracting each other:
By above-mentioned formula, calculate the difference image of present frame and a rear frame, and calculate according to this difference image of present frame and former frame.
(C) centered by (x, y), get 3 * 3 moving window, N η=9, (x, y) some Fourth-order moment mean value
Wherein
mean value for frame-to-frame differences gray-scale value in moving window:
(D) binaryzation
Choose appropriate threshold, for three-frame difference method, choose rational threshold value the effect of target extraction is played an important role, with optimizing, strengthen large Tianjin method definite threshold below.
The 3rd step: in second step, threshold value chooses.Large Tianjin method after utilize improving is that particle group optimizing strengthens large Tianjin method and determines optimal threshold.Strengthen large Tianjin method owing to introducing the time that has also increased computing when gray scale stretching has improved large Tianjin method effect, when the threshold value that is applied to real-time online is chosen, be unfavorable for true-time operation, time-based requirement, the present embodiment improves strengthening large Tianjin method, adopt particle swarm optimization algorithm optimization to strengthen large Tianjin method, first to the position of particle and speed, give a RANDOM SOLUTION, owing to being that gray level image scope is between 0-255, so will be within the scope of this given particle position, all random values are produced by rand () function.Then calculate fitness function value and compare, iterative computation, finally tries to achieve optimum solution, namely optimal threshold.
Weight w and c in computation process
1, c
2isoparametricly choose very crucially, wherein the selection of weight w affects the search capability of method.Generally, wish has stronger ability of searching optimum in initial, and stage is laid particular stress on the enhancing of local search ability; Generally in search procedure, w value, from 0.9 linear decrease to 0.4, in order to make particle can find better optimal value, is generally got c
1=c
2=2.
The 4th step: according to the optimal threshold in the formula in the first step (6) and second step, difference result is carried out to binaryzation, determine binary image.
The 5th step: the imagery exploitation image in (5) is processed to post-processing technology, as: 1) first with less structural element B1, image is done once to " opening " computing (Open Operation) and eliminate isolated noise; 2) with larger structural element B2, image is done once to " closing " computing (Close Operation) again and fill up target internal cavity; Further improve incomplete border, thereby obtain complete boundary image, thereby realize the smooth extraction of flco target.
The 6th step: software emulation is realized, and Fig. 3 is simulation result.By common 3 kinds of methods and the inventive method, completed respectively the design and experiment of motion detection block, whole program is based on Windows operating system, and having adopted the matlab7.0 of mathwork company or the Visual C++6.0 of Microsoft is programming development platform.
The present embodiment shooting machines configurations used is CCD Techno-Industrial video camera, used 10 groups of video sequence images of taking to test, each sequence contains 30 frame left and right, image size is 312 * 288 tiff form, 256 * 190 gif format-pattern etc., get respectively representational continuous three two field pictures, with common 3 kinds of methods (two three-frame difference separating methods, three three-frame difference separating method+alternative manners, three three-frame differences divide+large Tianjin method) and method of the present invention, completed respectively the design and experiment of motion detection block, whole program is based on Windows operating system, having adopted the matlab7.0 of mathwork company and the Visual C++6.0 of Microsoft is programming development platform.
Claims (3)
1. a three-frame difference high-order statistic is in conjunction with the flco detection method of OTSU algorithm, it is characterized in that, described method is irregular based on motion flco, large reason affected by environment, use strengthens large Tianjin method based on three-frame difference, high-order statistic and particle group optimizing and extracts floc characteristic, comprises the following steps:
(1) respectively to present frame I
t, former frame I
t-1, a rear frame I
t+1carry out pre-service, comprise gray scale correction, the processing such as figure image intensifying;
(2) I
t-1, I
t, I
t+1do respectively difference, obtain difference image D
t+1, D
t;
(3) difference image is carried out obtaining difference image D with computing
i;
(4) utilize particle group optimizing to strengthen large Tianjin method and determine optimal threshold T;
(5) by pixel calculate Fourth-order moment and with optimal threshold T comparison, be greater than T value, think that it is moving target, pixel value is set to 1, otherwise is 0;
(6) to the image in step (5), utilize Image Post-processing Techniques, further improve incomplete border, thereby obtain complete boundary image, thereby realize the smooth extraction of flco target.
2. a kind of three-frame difference high-order statistic according to claim 1 is in conjunction with the flco detection method of OTSU algorithm, it is characterized in that, described definite optimal threshold, first to the position of particle and speed, give a RANDOM SOLUTION, owing to being that gray level image scope is between 0-255, so will be within the scope of this given particle position, all random values are produced by rand () function; Then calculate fitness function value and compare, iterative computation, finally tries to achieve optimum solution, namely optimal threshold.
3. a kind of three-frame difference high-order statistic according to claim 1, in conjunction with the flco detection method of OTSU algorithm, is characterized in that, described Image Post-processing Techniques is first done once " opening " computing with less structural element to image and eliminated isolated noise; With larger structural element, image is done once to " closing " computing again and fill up target internal cavity; Further improve incomplete border, thereby obtain complete boundary image, thereby realize the smooth extraction of flco target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310658750.3A CN103632373B (en) | 2013-12-09 | 2013-12-09 | A kind of flco detection method of three-frame difference high-order statistic combination OTSU algorithms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310658750.3A CN103632373B (en) | 2013-12-09 | 2013-12-09 | A kind of flco detection method of three-frame difference high-order statistic combination OTSU algorithms |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103632373A true CN103632373A (en) | 2014-03-12 |
| CN103632373B CN103632373B (en) | 2017-12-22 |
Family
ID=50213390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310658750.3A Active CN103632373B (en) | 2013-12-09 | 2013-12-09 | A kind of flco detection method of three-frame difference high-order statistic combination OTSU algorithms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103632373B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997598A (en) * | 2017-01-06 | 2017-08-01 | 陕西科技大学 | The moving target detecting method merged based on RPCA with three-frame difference |
| CN109828236A (en) * | 2019-02-14 | 2019-05-31 | 中南大学 | A kind of microseism/acoustic emission source locating method in labyrinth containing dead zone |
| CN111833269A (en) * | 2020-07-13 | 2020-10-27 | 字节跳动有限公司 | Video noise reduction method and device, electronic equipment and computer readable medium |
| CN112418105A (en) * | 2020-11-25 | 2021-02-26 | 湖北工业大学 | High maneuvering satellite time sequence remote sensing image moving ship target detection method based on difference method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184550A (en) * | 2011-05-04 | 2011-09-14 | 华中科技大学 | Mobile platform ground movement object detection method |
-
2013
- 2013-12-09 CN CN201310658750.3A patent/CN103632373B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184550A (en) * | 2011-05-04 | 2011-09-14 | 华中科技大学 | Mobile platform ground movement object detection method |
Non-Patent Citations (2)
| Title |
|---|
| 周红娟: "粒子群优化增强大津法的气泡分割方法研究", 《东北电力大学学报》 * |
| 张之稳: "嵌入式视频跟踪算法的研究", 《中国优秀硕士学位论文全文数据库》, 15 December 2006 (2006-12-15), pages 4 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106997598A (en) * | 2017-01-06 | 2017-08-01 | 陕西科技大学 | The moving target detecting method merged based on RPCA with three-frame difference |
| CN109828236A (en) * | 2019-02-14 | 2019-05-31 | 中南大学 | A kind of microseism/acoustic emission source locating method in labyrinth containing dead zone |
| CN111833269A (en) * | 2020-07-13 | 2020-10-27 | 字节跳动有限公司 | Video noise reduction method and device, electronic equipment and computer readable medium |
| CN111833269B (en) * | 2020-07-13 | 2024-02-02 | 字节跳动有限公司 | Video noise reduction method, device, electronic equipment and computer readable medium |
| CN112418105A (en) * | 2020-11-25 | 2021-02-26 | 湖北工业大学 | High maneuvering satellite time sequence remote sensing image moving ship target detection method based on difference method |
| CN112418105B (en) * | 2020-11-25 | 2022-09-27 | 湖北工业大学 | High maneuvering satellite time sequence remote sensing image moving ship target detection method based on difference method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103632373B (en) | 2017-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104392468B (en) | Based on the moving target detecting method for improving visual background extraction | |
| CN102307274B (en) | Motion detection method based on edge detection and frame difference | |
| CN104834912B (en) | A kind of weather recognition methods and device based on image information detection | |
| CN105427626B (en) | A kind of statistical method of traffic flow based on video analysis | |
| CN103136766B (en) | A kind of object conspicuousness detection method based on color contrast and color distribution | |
| CN104268872B (en) | Consistency-based edge detection method | |
| CN103955949B (en) | Moving target detecting method based on Mean-shift algorithm | |
| CN103258332B (en) | A kind of detection method of the moving target of resisting illumination variation | |
| CN112215819B (en) | Airport pavement crack detection method based on depth feature fusion | |
| CN103632373A (en) | Floc detection method combining three-frame differential higher-order statistics (HOS) with OTSU algorithm | |
| CN104408741A (en) | Video global motion estimation method with sequential consistency constraint | |
| CN103679677A (en) | Dual-model image decision fusion tracking method based on mutual updating of models | |
| CN105787912A (en) | Classification-based step type edge sub pixel localization method | |
| CN104700405A (en) | Foreground detection method and system | |
| CN112580661A (en) | Multi-scale edge detection method under deep supervision | |
| CN105389820A (en) | Infrared image definition evaluating method based on cepstrum | |
| CN105205834A (en) | Target detection and extraction method based on Gaussian mixture and shade detection model | |
| CN104835142A (en) | Vehicle queuing length detection method based on texture features | |
| CN107871315B (en) | Video image motion detection method and device | |
| CN107358635B (en) | Color morphological image processing method based on fuzzy similarity | |
| CN105139358A (en) | Video raindrop removing method and system based on combination of morphology and fuzzy C clustering | |
| CN105741317B (en) | Infrared motion target detection method based on time-space domain significance analysis and rarefaction representation | |
| CN108009480A (en) | A kind of image human body behavioral value method of feature based identification | |
| CN105844671B (en) | A kind of fast background relief method under the conditions of change illumination | |
| Xiang et al. | Research on target detection method based on HSV fusion Gaussian mixture model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |