CN111462143A

CN111462143A - Watershed algorithm-based insect body recognition and counting method and system

Info

Publication number: CN111462143A
Application number: CN202010205190.6A
Authority: CN
Inventors: 张洪; 马尧辉; 林俊宇
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2020-03-22
Filing date: 2020-03-22
Publication date: 2020-07-28
Anticipated expiration: 2040-03-22
Also published as: CN111462143B

Abstract

The invention discloses a watershed algorithm-based insect body recognition and counting method and system, and belongs to the technical field of image processing. The method of the invention comprises the following steps: collecting an image containing a worm body, and carrying out binarization processing to obtain a binary image; deleting the contour of which the area is smaller than the minimum value of the worm body area in the binary image, and filling the worm body contour to obtain a filled image; gradually corroding the filled image, marking the outlines of all worm bodies, and performing expansion processing on the marked background to obtain a first marked image; further processing the filled image and the first marker map by using a watershed algorithm to obtain a second marker map; and acquiring the roundness and the area of the contour in the second marking map, and counting the number of the contours to finish counting. Compared with the traditional mode of counting by manpower, the method can automatically count the number of the insect bodies, has higher efficiency and recognition counting accuracy, and is suitable for counting various insect pests.

Description

Watershed algorithm-based insect body recognition and counting method and system

Technical Field

The invention belongs to the technical field of image processing, and particularly relates to a method and a system for identifying and counting insect bodies based on a watershed algorithm.

Background

In China, crop losses caused by plant diseases and insect pests are quite remarkable every year, so that the prevention and control of the plant diseases and insect pests have important significance for guaranteeing the economic development of the nation, and the prevention and control time is judged and the prevention and control effect is evaluated when the occurrence degree of the plant diseases and insect pests is judged.

Taking the tobacco industry as an example, the tobacco pests are the main pests in the tobacco industry in China, and in the whole tobacco circulation and production process, the tobacco pests can be harmed in each link from tobacco leaf purchase to threshing and redrying, storage and transportation, production and processing, and finished product storage, transportation and sale. Traps consisting of a piece of cardboard coated with adhesive and a trap core are important tools for monitoring the presence of smoke insects, which stick when attracted by the smell of the trap core. Currently, the density of the tobacco insects in the front area is generally known by manually counting the number of the insects on the trap, and therefore when the number of the tobacco insects is large, the surrounding environment is timely cleaned and killed. On one hand, however, the manual counting mode is difficult to realize on some special smoke worm monitoring points, for example, the smoke worm counting at the top end of an elevated warehouse requires workers to climb onto shelves with several floors of height, so that certain potential safety hazards exist; there are also some smoke insect monitoring points in the air-conditioning pipeline, and workers are difficult to enter the checking. On the other hand, when the number of the tobacco worms is large, a certain counting error exists in the manual counting mode, and the efficiency is low.

Disclosure of Invention

Aiming at the defects of the prior art, the invention aims to provide a watershed algorithm-based method and a watershed algorithm-based system for identifying and counting insect bodies, and aims to solve the problems that the manual counting mode is low in accuracy and efficiency and has potential safety hazards sometimes.

In order to achieve the above object, the present invention provides an insect recognition and counting method based on watershed algorithm, including the following steps:

collecting an image containing a worm body, and carrying out binarization processing to obtain a binary image;

deleting the contour of which the area is smaller than the minimum value of the worm body area in the binary image, and filling the worm body contour to obtain a filled image;

gradually corroding the filled image, marking the outlines of all worm bodies, and performing expansion processing on the marked background to obtain a first marked image;

further processing the filled image and the first marker map by using a watershed algorithm to obtain a second marker map;

and acquiring the roundness and the area of the contour in the second marking map, and counting the number of the contours to finish counting.

Wherein the binarization processing further includes:

and (4) respectively separating the RGB color space and the L AB color space, and removing interference factors according to the characteristics of different channel images.

Further, gradually eroding the filled image, marking all the worm body outlines, and performing expansion processing on the marked background to obtain a first marked image comprises:

newly building a binary image, processing the filled image by adopting a gradual corrosion method, setting an area threshold, and marking the position of the contour on the newly built binary image when the area of the contour of the worm is smaller than the area threshold;

and performing expansion processing, and marking a background on the newly-built binary image.

Furthermore, when the number of the contours is counted, all the worm body contours are traversed, the contours with the roundness not meeting the standard and the areas more than one time larger than the average area are considered as the contours formed by adhesion of a plurality of worm bodies, and the number of the worm bodies contained in the contours is determined by an area comparison method.

Further, after counting is completed, according to the distribution of the outlines in the second label map, the outlines are filled with random colors, and the information of the total number of the insects is displayed.

The invention also provides a watershed algorithm-based insect body recognition and counting system, which comprises a collecting unit and a processing unit; wherein

The acquisition unit is used for acquiring an image containing a worm body and carrying out binarization processing to obtain a binary image;

the processing unit includes:

the filling unit deletes the contour of which the area is smaller than the minimum value of the worm body area in the binary image, and fills the worm body contour to obtain a filled image;

the first marking unit is used for gradually corroding the filled image, marking the outlines of all worm bodies, and performing expansion processing on the marked background to obtain a first marked image;

the second marking unit is used for further processing the filled image and the first marking image by utilizing a watershed algorithm to obtain a second marking image;

and the counting unit is used for acquiring the roundness and the area of the outline in the second marking map, and counting the number of the outlines to finish counting.

Further, the binarization processing further includes:

Furthermore, when the number of the contours is counted, the counting unit traverses all the contours of the worm bodies, the contours with the roundness not meeting the standard and the areas more than one time larger than the average area are regarded as the contours formed by adhesion of a plurality of worm bodies, and the number of the worm bodies contained in the contours is determined by an area comparison method.

Further, the system also comprises

And the display unit is used for filling the contour by using random colors according to the contour distribution in the second marking map after counting is finished, and displaying the information of the total number of the insects.

Compared with the prior art, on one hand, the insect body recognition counting method provided by the invention has the advantages that the defects of high cost, potential safety hazard, low efficiency and the like of the existing manual cigarette counting method can be improved by acquiring the insect body image and then automatically counting the image; on the other hand, on the basis of removing the interference factors in the picture and carrying out filling processing, the image recognition method based on the watershed algorithm is adopted, so that the counting accuracy is improved, and the method is suitable for counting various insects.

Drawings

FIG. 1 is a flowchart of an insect recognition and counting method based on a watershed algorithm according to an embodiment of the present invention.

Fig. 2 is a flowchart of a method for identifying and counting tobacco insects based on a watershed algorithm according to an embodiment of the present invention.

Figure 3 is a diagram of a tobacco insect trap taken in the field of a cigarette factory in an embodiment of the present invention.

Fig. 4 is a cut-out view of the original image without black frame interference.

Fig. 5(a) is a diagram showing the result of adaptive binarization of a G-channel image, fig. 5(B) is a L-channel binary image, fig. 5(c) is an a-channel binary image, and fig. 5(d) is a B-channel binary image.

Figure 6 is a binary diagram with tobacco, trap and spot interference removed.

FIG. 7 is a binary image processed by filling and deleting small area contours.

FIG. 8 is a binary image with background lines removed by an open operation.

Fig. 9 is a labeled graph after a gradual erosion and expansion treatment.

FIG. 10 is a marker map after processing by the watershed algorithm.

Fig. 11 is a graph of the final display result obtained by the method of the present embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

As shown in fig. 1, an embodiment of the present invention provides an insect body identification and counting method based on a watershed algorithm, including the following steps:

acquiring the roundness and the area of the outline in the second marking map, counting the number of the outlines and counting;

and after counting is finished, filling the contour with random colors according to the contour distribution in the second labeled graph, and displaying the information of the total number of the insects.

Wherein the binarization processing further includes:

As can be understood by those skilled in the art, the method for identifying and counting the insects based on the watershed algorithm provided by the invention can be applied to statistics of various insects. Taking the situation of the statistics of the tobacco insects as an example, a method for identifying and counting the tobacco insects based on the watershed algorithm is introduced, and a flow chart is shown in fig. 2, and the method comprises the following steps:

(1) shooting images of the tobacco worm trapper by the camera, and intercepting an effective tobacco worm detection area to obtain a shearing graph;

(2) separating RGB color space and L AB color space of the cut picture respectively, and removing interference factors such as tobacco leaves, trapping cores, dust and the like in the original picture according to the characteristics of different channel pictures;

(3) deleting the outline with the area smaller than the minimum value of the area of the tobacco worm in the picture, and filling the outline of the tobacco worm to ensure that the outline is complete;

(4) removing the background line of the tobacco pest trapper by using an opening operation;

(5) marking all the smoke worm outlines by using a gradual corrosion method, and carrying out expansion treatment to mark a background;

(6) further processing the marker map by using a watershed algorithm to obtain a watershed algorithm marker map;

(7) calculating the roundness and the area of the contour in the watershed algorithm mark image, and counting the number of the contours;

(8) all contours are filled in the cut map with random colors according to the contour distribution in the watershed algorithm marker map, and the total number of smoke worms is shown in the upper left corner of the cut map.

Preferably, in the step (1), the picture of the tobacco pest trap is shot through the camera, and an effective tobacco pest monitoring area needs to be intercepted to obtain a cut-off graph because the shot picture may further include other interference objects such as a wall surface and the like.

Preferably, in the step (2), the cut image obtained in the step (1) is separated into an RGB color space and an L AB color space respectively, tobacco leaves are generally yellow, a trap core installed on the trap is generally red or white, and objects which are not smoke and insects can be identified through the separated color space, a G channel image is selected to be subjected to self-adaptive binarization processing to obtain a binary image of the cut image, the L channel image can be subjected to binarization processing to remove facula interference caused by flash lamp reflection in the binary image through contrast filling operation, the A channel image can be subjected to binarization processing to remove interference caused by the trap core in the binary image through contrast filling operation, the B channel image can be subjected to binarization processing to remove interference caused by interference factors such as tobacco leaves in the binary image through contrast filling operation, and the above specific processing can be adjusted according to actual needs when the cut image is used for identification and counting of other insects.

Preferably, in the step (3), the interference-removed binary image obtained in the step (2) is subjected to contour searching operation, and area information of all contours is calculated. And setting a threshold value of the area of the tobacco worm, filling the outline into black when the area of the outline is smaller than the threshold value, and removing the outline. And simultaneously calculating the center distance information of all the contours, and filling white if the center distance position of the contours is black so as to complete the contours of the tobacco insects.

Preferably, in the step (4), if the smoke insect trap carries a background line, the contour obtained in the step (3) is filled with a complete binary image according to actual needs to perform an open operation, so as to remove the background line carried by the smoke insect trap.

Preferably, in step (5), a binary image is created, the binary image processed in step (4) is processed by using a gradual erosion method, an area threshold is set, and when the area of the contour is smaller than the set value, the position of the contour is marked on the created binary image. And (5) similarly, performing expansion processing on the binary image processed in the step (4), and marking a background on the newly-built binary image.

Preferably, in step (6), the new marker map obtained in step (5) and the binary map obtained in step (4) are input into a watershed algorithm to obtain a watershed algorithm marker map.

Preferably, in the step (7), the roundness and area information of the contour of the watershed algorithm marker map obtained after the step (6) are calculated, and the number of the contour is counted. And traversing all the contours, regarding the contour with the roundness not meeting the standard and the area more than one time larger than the average area as the contour formed by the adhesion of a plurality of tobacco insects, and determining that a plurality of tobacco insects exist in the contour by an area comparison method.

Preferably, in step (8), in the cut map obtained in step (1), the contours are filled with random colors according to the distribution of the contours in the watershed algorithm marker map, and information on the total number of the tobacco worms is displayed in the upper left corner of the cut map.

The embodiment adopts the watershed algorithm-based tobacco worm identification and counting method provided by the invention to verify on a field-shot tobacco worm trap picture cooperated with a cigarette factory, the picture used for the embodiment is shot on the field of the cigarette factory as shown in fig. 3, an original picture is cut to remove the interference of a surrounding black frame as shown in fig. 4, fig. 5(a) is a result of performing RGB color space separation on the cut picture and then taking a G channel image for self-adaptive binarization, fig. 5(B), fig. 5(c) and fig. 5(d) are respectively a L channel image, an A channel image and a B channel image which are subjected to binarization processing after performing L AB color space separation on the cut picture, and the binary image with three channels is compared and filled in the G channel binary image, so that the binary image with the interference of tobacco leaves, a trap core and light spots removed is shown in fig. 6.

And carrying out contour searching operation on the interference-removed binary image, and calculating the area information of all contours. And setting a threshold value of the area of the tobacco worm, and filling the outline into black when the area of the outline is smaller than the threshold value, and removing the outline. And simultaneously calculating the center distance information of all the contours, and filling white if the center distance position of the contour is black so as to complete the contour of the tobacco worm, wherein the result is shown in figure 7.

Fig. 8 is a binary image in which the background line is removed after the opening operation is performed. Fig. 9 is a marker map in which the binary map shown in fig. 8 is subjected to a gradual erosion process on the contour and an expansion process on the background. With the background marked black. FIG. 10 is a watershed algorithm marker map processed using a watershed algorithm, which appears to the naked eye as a black image because the contour pixel values of the watershed algorithm marker start at 1.

And (4) calculating the roundness and area information of the contour of the watershed algorithm marker map, and counting the number of the contours. And traversing all the contours, regarding the contour with the roundness not meeting the standard and the area more than one time larger than the average area as the contour formed by the adhesion of a plurality of tobacco insects, and determining that a plurality of tobacco insects exist in the contour by an area comparison method. In the cut-out map, the contours are filled with random colors according to the contour distribution in the watershed algorithm labeled map, and the information of the total number of the tobacco insects is displayed in the upper left corner of the cut-out map, and the display result is shown in fig. 11.

Correspondingly, the embodiment of the invention also provides an insect body recognition and counting system based on the watershed algorithm, which comprises a collecting unit and a processing unit; wherein

the processing unit includes:

the counting unit is used for acquiring the roundness and the area of the outline in the second marking map, counting the number of the outlines and counting;

Further, the binarization processing further includes:

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. An insect body recognition and counting method based on a watershed algorithm is characterized by comprising the following steps:

2. The worm recognition and counting method according to claim 1, wherein the binarization processing further comprises:

3. The worm recognition and counting method of claim 1, wherein gradually eroding the filled image, marking all worm contours, and expanding the marked background to obtain the first marked image comprises:

4. The worm recognition and counting method of any one of claims 1 to 3, wherein when counting the number of the contours, all worm contours are traversed, contours with roundness not meeting the standard and area more than one time the average area are considered as the contours formed by adhesion of a plurality of worms, and the number of the worms contained in the contours is determined by the area comparison method.

5. The worm recognition and counting method of any one of claims 1 to 3, wherein after counting is completed, the contour is filled with random colors according to the contour distribution in the second label chart, and information on the total number of the worms is displayed.

6. An insect body recognition and counting system based on a watershed algorithm is characterized by comprising a collecting unit and a processing unit; wherein

the processing unit includes:

7. The worm recognition counting system of claim 6, characterized in that the binarization process further comprises:

8. The worm recognition and counting system of claim 6, wherein the gradually eroding the filled image, marking all worm contours, and the expanding the marked background to obtain the first marked map comprises:

9. The worm recognition and counting system of any one of claims 6 to 8, wherein the counting unit is configured to traverse all worm contours during counting the number of the contours, consider the contours with roundness not meeting the standard and area more than one time larger than the average area as the contours formed by adhesion of a plurality of worms, and determine the number of the worms contained in the contours by area comparison.

10. The worm recognition and counting system of any one of claims 6 to 8, further comprising