CN109961445A

CN109961445A - A kind of watermelon quirk character determination method

Info

Publication number: CN109961445A
Application number: CN201910236205.2A
Authority: CN
Inventors: 不公告发明人
Original assignee: Individual
Current assignee: Individual
Priority date: 2019-03-27
Filing date: 2019-03-27
Publication date: 2019-07-02

Abstract

The invention discloses a kind of watermelon quirk character determination methods, belong to field of image recognition this method to be first split watermelon quirk and the extraction of target area, the subscale for obtaining passing through after target area hsv color space extracts to obtain green area to the green area in image, and the judgement of watermelon quirk character is carried out by green area area, rapidly and accurately watermelon quirk character can be determined.

Description

A kind of watermelon quirk character determination method

Technical field

The present invention relates to a kind of watermelon quirk character determination methods, specifically carry out mesh to the watermelon quirk of acquisition first The extraction in region, then the method that the green area area by extracting determines character are marked, field of image recognition is belonged to.

Background technique

China is the maximum country of growth of watermelon area, as China's industrialized agriculture constantly develops, the anniversary kind of watermelon The environment opposing seal of cultivation is planted, facility internal environment is more demanding to temperature and humidity, and facility plantation is especially embodied in greenhouse cultivation, Closed planting environment has completely cut off the entrance of extraneous insect pollinator, and watermelon is entomophilous flower, and insect is mainly leaned under natural conditions Pollination, therefore be mostly now that human assistance bears fruit (artificial pollination, growth regulator) to ensure the fruit-setting rate of watermelon, with machine For vision in extensive use agriculturally, intelligent watermelon pollination robot certainly will be research tendency, however intelligent pollination machine It is exactly the position for identifying watermelon flower that the key technology of people, which is selected, determines the character of quirk, however is obtaining the image of watermelon quirk When, since the watermelon grown in its natural state spends the interference that can inevitably have green, tendril and film, it is desirable to obtain background Single target quirk region needs to be split processing to the watermelon quirk image of acquisition, then identifies the character of quirk, mesh The character determination method of the preceding dividing method applied not yet in watermelon quirk image and watermelon quirk.

Summary of the invention

Effectively identification character effectively cannot be carried out to watermelon quirk image for the prior art, the present invention provides one Kind watermelon quirk character determination method, the green area by extracting segmented image can be fast and accurately to the property of watermelon quirk Shape is determined.

The present invention is achieved by the following technical solutions: a kind of watermelon quirk character determination method, this method are right first Watermelon quirk is split the extraction with target area, obtains passing through the subscale in hsv color space to figure after target area Green area as in extracts to obtain green area, and the judgement of watermelon quirk character is carried out by green area area, The following steps are included:

(1) image of the rgb space of acquisition is filtered first, is transformed into Lab color space and in the channel b benefit Be split with OTSU algorithm, negated and Morphological scale-space to binary map is obtained, finally will treated binary map and original image with Color Segmentation figure is obtained after operation.

(2) HSV space is transformed into Color Segmentation effect picture, takes HSV green component table to carry out green to image by looking into Region extracts, and determines watermelon quirk character by obtained green area area.

The pretreatment operation that described image carries out is median filter process, and the convolution kernel size for planting filtering is 5*5.

Described that the bianry image that Otsu automatic threshold segmentation obtains is carried out to the channel b, target area color is black, back Scape is white, Ying Jinhang inversion operation, while carrying out Morphological scale-space to image, and Morphological scale-space includes area filling, expansion The operation such as corrosion and the removal of small area region.

The binary map is single-pass image, should be mapped to triple channel and carry out and operate with original image again.

The Color Segmentation effect picture be transformed into HSV space then can extract maximum green region area have greater than etc. It is female flower in 2000, the maximum green region area of extraction is male flower less than 2000.

The usefulness of the invention is can be split processing to the watermelon quirk image of acquisition, rapidly and accurately divide It cuts and extracts watermelon quirk target area, extract green area by extracting in Color Segmentation image, can accurately determine watermelon The character of quirk.The method use image of the median filtering function to acquisition to be pre-processed, the core of median filtering function The 5*5 size that function uses retains the marginal information of image, in addition by acquisition while capable of effectively removing noise jamming The image of RGB color is transformed into Lab color space, carries out threshold value point to the channel b using Otsu automatic threshold segmentation algorithm The bianry image cut can carry out primary segmentation to image, eliminate a large amount of background interference；Primary segmentation is obtained Bianry image carry out Threshold segmentation, by morphologic holes filling and small area removal operation after, obtain background it is single two It is worth image；For bianry image inversion operation, and bianry image be it is single pass be mapped to triple channel, can be carried out with original image with Operation obtains Color Segmentation effect picture, and obtained colour picture only has watermelon quirk region, can be preferably to watermelon quirk Region carries out green area extraction, eliminates the interference of green and tendril, can more accurately determine the character of watermelon quirk.

Detailed description of the invention

Fig. 1 is the flow chart of watermelon quirk character method of discrimination；

Fig. 2 is the effect picture of image median filter；

Fig. 3 is watermelon quirk Color Segmentation effect picture；

Fig. 4 is hsv color subscale；

Fig. 5 is that watermelon quirk green areas extracts schematic diagram.

Specific embodiment

A kind of watermelon quirk character determination method, as shown in Figure 1, this method is split first to watermelon quirk and target The extraction in region, the subscale for obtaining passing through after target area hsv color space extract the green area in image Green area is obtained, and carries out the judgement of watermelon quirk character by green area area, specifically includes the following steps:

(1) image of acquisition is pre-processed first as shown in Figure 2, pretreatment operation is median filter process, and is planted Planting filtering processing using convolution kernel is 5*5 size, and median filter process can effectively remove the same of noise jamming When retain image marginal information.

(2) since the image of acquisition is RGB color, image is transformed into Lab color space, and return Lab space Image carries out channel separation operation, obtains b channel components figure, carries out the binary map that Otsu automatic threshold segmentation obtains to the channel b Picture, since target area color is black in bianry image, background is white, should be subtracted into the pixel for traversing whole sub-picture, use 255 Each pixel value is removed, so that its value pixel is 0 to take 255, the pixel for being 255 to its pixel value takes 0, completes taking for image Inverse operations, so that target area is shown as white, background area is black, finally obtained bianry image such as Fig. 3 secondary series Shown in image.

(3) since Fig. 3 secondary series bianry image is there are other interference such as white point, hole, morphology need to be carried out to it Processing, basic removal interference, obtains the binary map as described in arranging Fig. 3 third, and wherein Morphological scale-space includes area filling, expansion Corrosion and small area region remove processing operation, and the binary map that aforesaid operations obtain is single channel image, need to be mapped to threeway Road is carried out and is operated with original image (RGB triple channel) again, obtains color images effect picture, the effect shown in the column of Fig. 3 the 4th Fruit figure, can obtain the color image of watermelon quirk target area.

(4) it can be seen that watermelon male flower and female Huadu have the flower region of yellow, mainly from the image after segmentation Difference is below female flower there are ovary, and the color of ovary is green, therefore by the target color figure of obtained watermelon quirk As being transformed into hsv color space, then looks into and take hsv color subscale, its green area is extracted, although the image of front The interference of greenery and tendril has been had been removed in cutting procedure, but there are other green interferences in inevitable image again, it is right The green areas region of extraction carries out area judgement, and maximum green area area is male flower less than 2000, maximum green It is female flower that color area, which is greater than 2000,.

For the ordinary skill in the art, introduction according to the present invention, do not depart from the principle of the present invention with In the case where spirit, changes, modifications that embodiment is carried out, replacement and variant still fall within protection scope of the present invention it It is interior.

Claims

1. a kind of watermelon quirk character determination method, it is characterised in that: this method is split first to watermelon quirk and target The extraction in region, the subscale for obtaining passing through after target area hsv color space extract the green area in image Green area is obtained, and green area area is determined, comprising the following steps:

(1) image of the rgb space of acquisition is pre-processed first, be transformed into Lab color space and utilize OTSU in the channel b Algorithm is split, and is negated and Morphological scale-space to binary map is obtained, finally will be after treated binary map and original image and operation Obtain Color Segmentation figure；

(2) HSV space is transformed into Color Segmentation effect picture, takes HSV green component table to carry out green area to image by looking into It extracts, watermelon quirk character is determined by obtained green area area.

2. a kind of watermelon quirk character determination method according to claim 1, it is characterised in that: described image carries out pre- Processing operation is median filter process, and the convolution kernel size for planting filtering is 5*5.

3. a kind of watermelon quirk character determination method according to claim 1, it is characterised in that: described to be carried out to the channel b The bianry image that Otsu automatic threshold segmentation obtains, target area color are black, and background is white, and Ying Jinhang negates behaviour Make, while Morphological scale-space is carried out to image, Morphological scale-space includes area filling, dilation erosion and the removal of small area region etc. Operation, the binary map are single-pass image, should be mapped to triple channel and carry out and operate with original image again.

4. a kind of watermelon quirk character determination method according to claim 3, it is characterised in that: the Color Segmentation effect Fruit figure is transformed into HSV space, and it is female flower, the maximum green of extraction more than or equal to 2000 that the maximum green region area of extraction, which has, Region area is male flower less than 2000.