CN102955935A - Sediment recognition method and sediment recognition system based on image fusion - Google Patents

Abstract

Disclosed are a sediment recognition method and a sediment recognition system based on image fusion. The sediment recognition system is a novel digital recognition system for sediment particle size and proportion. Natural sediment is recognized completely by information processing and artificial intelligence decision technology. The sediment recognition method includes: shooting to-be-recognized sediment, and processing according to the shot pictures. The sediment recognition method and the sediment recognition system are modified. The sediment recognition system comprises a multi-dimensional image data acquisition part, a multi-dimensional image data fusion part and a sediment particle size and proportion recognition and intelligent decision part. Sediment particle size and proportion are recognized on the basis of image fusion due to three-dimensional inherent attribute, multi-dimensionality and irregularity of sediment and according to image processing technology, especially technology for making two-dimensional images three-dimensional.

Description

The method and system of the silt identification of merging based on image

Technical field

The present invention relates to the cross discipline of hydro science and information science.

Background technology

Based on the importance of sand grains gradation, grain composition how to obtain required silt becomes the basic problem in science of an outbalance.China's River Hydrology circle has used respectively particle diameter meter method, pipette method, light extinction method since nineteen fifty is carried out Particle Size Analysis of River Load work, the Particle Size Analysis of River Load method that sieve method etc. are traditional.Recent years, progressively formed again a kind of comparatively advanced method---laser method.Now use instrument, step and the relative merits etc. of the method for various test sand grains gradations are done brief discussion.

Tape measurement: required instrument mainly contains bolter, vernier caliper, scale etc.At first the silt sample is separated large particle diameter silt with bolter, measured one by one three shaft lengths of bulky grain silt with vernier caliper again, be respectively a, b, c, then the particle diameter of this sand grain isThe weight that weighs up again each silt can obtain the gradation composition of silt.Tape measurement is applicable to particle diameter and measures greater than the sand grains gradation of 100mm, and major advantage is that equipment is simple, and visual result is easy to operate, and is only applicable to single large particle diameter silt, and labour intensity is large, and low grade of efficient then is its weak point.

Volumetric method: the instrument that needs mainly contains water container, graduated cylinder etc.The method step and tape measurement are basic identical, only when measuring a certain sand grain particle diameter, be different from tape measurement, volumetric method is that sand grain to be measured is put into the container that is filled with water, and the water-head of container drew the volume of this silt with this before and after this sand grain of compute was put into again.Volumetric method also is applicable to particle diameter and measures greater than the sand grains gradation of 100mm, should be noted that during measurement that the water container diameter should not be greater than a times of particle diameter, and the water in the container can not splash or overflow.It is simple that the advantage of volumetric method mainly contains equipment, and visual result is easy to operate, and shortcoming is that test volume is little, and labour intensity is large, and efficient is on the low side etc.

Photogrammetry: the instrument that needs mainly contains high-definition digital camera or one in video camera and with reference to one on scale etc.When measuring, at first be placed on the silt sample to be measured with reference to scale, then with camera or video camera the silt sample is taken pictures, at last obtaining image is amplified, can measure the particle diameter of each sand grain according to the reference scale, and then calculate the grating value of this sample.Photogrammetry is applicable to the particle diameters such as coarse sand or cobble greater than the sand grain of 20mm, and that its advantage is is simple to operate, visual result, degree of accuracy are higher, and shortcoming is to measure the bottom sand grains gradation that is covered by top layer silt.

Sieve formula: the instrument that needs mainly contains bolter and scale etc.At first silt sample to be measured is sieved into some gratings group with the bolter in various apertures, the weight that weighs up combo silts at different levels with scale again can draw total grating value of this sample.Sieve formula is present most widely used sand grains gradation measuring method, be applicable to the sand grains gradation analysis of sediment grain size size between 0.05mm～100mm scope, it has that surveying instrument is simple, visual result, applicability reach by force the advantages such as the specimen amount is large, and size of mesh is variable, complex operation is affected by human factors the shortcomings such as larger.

Clear water sedimentation, uniformly mixed up settlement method, centrifugal settling method, laser method: the measuring method of these four kinds of sediment grain sizes all is only applicable to the grading analysis of 2mm and following sediment grain size, and bed load discharge is not had practical significance, thereby this paper does not do one by one and is described in detail.

Various sand grains gradation measuring methods are shown in Table 1.

Table 1

In various sand grains gradation method of testings mentioned above, all there is such or such inevitable defective in each method, such as tape measurement, volumetric method and Photographic technique are only applicable to the sand grains gradation identification of greater particle size, and tape measurement and volumetric method exist testing efficiency to hang down inferior defective simultaneously, the screening rule exists labour intensity large, the test result precision can't guarantee and the deficiency such as complex operation step, the indirect methods of measurement such as clear water sedimentation then can only be used for the measurement of superfine water suspended sediment particle, then can't be suitable for for the slightly large sand grain of particle diameter especially upper river pebble bed-load silt.And the Digital Image Processing that this paper will be narrated and artificial intelligence rule have been avoided the defective of above each method fully, be expected to become a kind of brand-new, more accurately, the method for efficiently sand grains gradation identification.

Existing technologies mainly contains the defective of the following aspects:

1, welding is wanted in traditional silt sampling, and mining process is accidents caused easily simultaneously.

2, the cost of sampling is very high, therefore can not sample on a large scale.

3, the mode of image fusion is not adopted in sand grain identification, seldom adopts the technology such as digital image processing, decision discernment yet.

4, existing digital image processing method is identified the size of sand grain, all is the process of complanation.Lack multidimensional and three-dimensional concept and technology.

Summary of the invention

The present invention seeks to the defective for present sand grain size identification method existence, proposed a kind of new sand grain size identification method and system.

The invention provides a kind of recognition methods of sand grain size, the method mainly may further comprise the steps:

According to the requirement of institute's sand grain size identification,

First step:

Digital filming is carried out in required area of sampling or sampled point, in the shooting process, dual mode is arranged, the one, camera lens remains on about 1 meter from the distance of silt; In addition 180 degree that carry out of same silt are taken, taken once every 30 degree, so, same sampling spot is taken six times.Can carry out like this sediment analysis and the identification of multidimensional.

Second step:

Image pre-service: carry out the gray scale processing to gathering the silt image, carry out sharpening and process, carry out simultaneously the statistics of gray scale and contrast.

The 3rd step:

6 pictures to same sampled point merge, and amalgamation mode can adopt Gauss to mate and alignment so.

The 4th step

The recognizer of sand grain size:

Algorithm input: ganmma controller matrix and figure image texture matrix, the silt image sample of training.

1, use the BP neural network algorithm to carry out learning training, the data of training are the sample data of silt image.

2, training is merged ganmma controller matrix and figure image texture matrix after finishing;

3, the BP algorithm behind the use learning training is identified new data.

Algorithm output: the ratio of sand grain size.

Description of drawings

Fig. 1 is the method and system structural representation of a kind of new sand grain size identification of the present invention.

Embodiment

The native system overall construction drawing as shown in Figure 1.

The overall system function is divided into three parts:

One, multidimensional image acquisition;

Two, the multidimensional image merges;

Three, the identification of sand grain size.

The image acquisition process generally determines the sample point that gathers, the scope of collection by the domain expert; Because adopt the mode of digital filming, cost is very low, so, collecting sample that can be a large amount of.

Image pre-service: when obtaining silt grating image, because the restriction of digital camera self pixel, the interference of external complex environment---even such as uneven illumination, the air low visibility, the reasons such as the artificial and mechanical factor of other in shooting, transmission and the processing procedure cause the silt grating that obtains image blurring unclear, and quality is on the low side, computing machine can't directly carry out discriminance analysis to it, and this just need to improve by certain preconditioning technique means the quality of silt grating image.General pretreated technological means mainly comprises image denoising, figure image intensifying and image recovery etc.

The image fusion: it is the core technology of this patent that image merges, and mainly is the samples of 180 degree samplings are carried out three-dimensional and merge.Here relate to the problems such as image aligning and coupling.Also relate to simultaneously the problem that flow pattern embeds.

The Intelligent Recognition of sand grain scale merges by the picture to 180 degree samplings at last, carries out the analysis of texture, obtains the ratio of rational grain size.

Claims (3)

1. According to the requirement of measured silt, system and practical application, use digital filming that the data object is gathered, so in fact native system comprises three sub-systems, claim is as follows:

   1. the system that protects the whole silt that merges based on image to identify
2. 2. protection is carried out pre-service to the silt image.Mainly comprise the systems such as image denoising, figure image intensifying and image recovery.
3. 3. protection is used Image Segmentation Using is processed, and then carries out feature extraction, and the result according to feature extraction classifies to image again, draws at last the result's of image recognition system.

Images