CN104949981A - Automatic detection method and system for cotton five-euphylla period - Google Patents

Abstract

The invention discloses a method for detecting the five true leaf stages of cotton, which comprises the following steps: (1) collecting an image of a single row of plants in a cotton field after seedling setting, splitting the image into sub-images of a single cotton plant, performing color segmentation and crop Image segmentation to obtain sub-images of cotton plants after seedling setting; (2) For all the sub-images of cotton plants, detect plant edges and skeletons to obtain plant main stems; (3) detect initial side stems of plants on both sides of the main stem position, The primary side stems with an acute angle on the upper side of the main stem in the images on both sides are used as side stems; (4) The intersection point of the side stem and the main stem is used as a node, when the number of sub-images with 2 or more nodes is detected to account for all When the number of subimages of cotton plants after seedling setting is more than 50%, it is considered that the cotton field has entered the five true leaf stage. The present invention also provides a system for realizing the above method. The detection result of the invention is accurate, the real-time performance is strong, automatic observation is realized, and manpower is saved.

Description

A method and system for automatic detection of five true leaf stages of cotton

technical field

The invention belongs to the intersection field of digital image processing and agricultural meteorological observation, and more specifically relates to a method and system for automatic detection of the five true leaf stages of cotton.

Background technique

Cotton is one of the main economic crops in China, and China's cotton output is also in the leading position in the world. The five true leaf stage of cotton is an important part of cotton growth and an important content of agricultural meteorological observation.

For a long time, the relevant information of the cotton development period has been recorded mainly by manual observation and recording. The observation results will be affected by the subjective factors of the observers, resulting in relatively large errors; at the same time, due to the long growth cycle of cotton, cotton planting The range is relatively wide, and the method of only using manual observation is time-consuming and labor-intensive.

Contents of the invention

In view of the above defects or improvement needs of the prior art, the present invention provides a method and system for automatic detection of the five true leaf stages of cotton, the purpose of which is to analyze the photos of cotton fields by image processing methods, thereby judging whether cotton has entered the five true leaf stages, by This solves the current time-consuming, labor-intensive and inaccurate technical problems of manually judging the five true leaf stages of cotton.

In order to achieve the above object, according to one aspect of the present invention, a method for automatic detection of the fifth true leaf stage of cotton is provided, which is characterized in that, comprising the following steps:

(1) Acquire sub-images of cotton plants after seedling setting: collect the front view of a single row of plants in a cotton field after seedling setting; use a search box suitable for the image size of a single cotton plant to split the image into sub-images with a certain split step Carrying out color segmentation and crop image segmentation to the sub-image, on the area segmentation result map of the crop segmentation method, retain the area occupied by the detected pixels in the color segmentation result map, and obtain the cotton plant sub-image after seedling setting;

(2) Detect the main stem of cotton plants: For all the sub-images of cotton plants obtained in step (1), the edge detection algorithm is used to detect the initial edge of the plant, and the skeleton detection algorithm is used to extract the initial skeleton of the plant. Chain code detection and straight line detection obtain the vertical plant edge and the plant skeleton of the cotton plant, and the inside of the vertical plant edge that contains the vertical plant skeleton is used as the main stem of the cotton plant to obtain the sub-image of the cotton plant that includes the main stem of the plant;

(3) Detect the side stems of cotton plants: Divide the sub-images of cotton plants obtained in step (2) containing the main stems into two sides according to the position of the main stems: the left side of the main stem is the left side of the image, and the right side of the main stem is the right side of the image; color segmentation and straight line detection are carried out on the images on both sides to obtain the primary lateral stems in the images on both sides, and the primary lateral stems which form an acute angle with the upper side of the main stem are used as lateral stems;

(4) Judgment of the five true leaf stages of cotton: For the cotton plant sub-image obtained in step (2) containing the main stem of the plant, the intersection of the side stem and the main stem is used as the node, when two or more nodes are detected When the number of sub-images accounts for more than 50% of the number of sub-images of cotton plants after seedling setting, it is judged that the cotton field has entered the five true leaf stage, otherwise, the next day's detection will be carried out.

Preferably, in the method for automatic detection of the five true leaf stages of cotton, the step (1) of the seedling setting time is judged according to the following steps: collect the lower view image of the cotton field under the same conditions every day, and use the segmentation method to perform green segmentation on the image , and counting the proportion of green pixels in the image is the green image coverage; compare the green image coverage of the lower view of the cotton field every day with the green image coverage of the lower view of the cotton field of the previous day, and when the green image coverage decreases, it is the seedling setting time .

Preferably, the automatic detection method of the five true leaf stages of cotton, the segmentation method used for the green segmentation, can adopt the environment adaptive segmentation method, the super green operator segmentation method, the crop image segmentation method based on Mean shift , Fisher linear discriminant method.

Preferably, in the automatic detection method of the five true leaf stages of cotton, the front view of the single row of plants is processed by a contrast stretching method.

Preferably, in the automatic detection method of the five true leaf stages of cotton, the search box is equal to the height of the front view of the single row of plants, and the width of the search box is 1/4 to 1/3 of its height. 2. The split step is 1/2 to 5/6 of the width of the search box.

Preferably, in the automatic detection method of the five true leaf stages of cotton, in the step (2) edge detection algorithm and skeleton detection algorithm, the image detection operators that can be used include Sobel operator, Roberts operator, LoG operator and Canny operator, preferably Canny operator.

Preferably, in the method for automatic detection of five true leaf stages of cotton, the straight line detection in the step (2) and step (3) can use Hough transform.

According to another aspect of the present invention, an automatic detection system for the five true leaf stages of cotton is provided, which is characterized in that it includes a cotton plant sub-image acquisition module, a cotton main stem detection module, a cotton side stem detection module, and a cotton five true leaf stage Judgment module;

The cotton plant sub-image acquisition module is used to collect the front view of a single row of cotton plants in a cotton field after seedling setting, split them into individual cotton plant sub-images, and process the sub-images into cotton plant sub-images after seedling setting to transfer to cotton main stem detection module;

The cotton main stem detection module is used to extract the edge of the cotton plant and its main skeleton, and the inner side of the vertical plant edge containing the vertical plant skeleton is used as the main stem of the cotton plant to obtain a sub-image of the cotton plant including the main stem of the plant, and The sub-image is delivered to the cotton lateral stem detection module;

The cotton side stem detection module is used to divide the cotton plant sub-image including the main stem of the plant into two sides according to the position of the main stem: the left side of the main stem is the left side of the image, and the right side of the main stem is the right side of the image; For the initial lateral stems in the images on both sides, the straight line forming an acute angle with the upper side of the main stem is used as the lateral stems, and the detection results are passed to the cotton five-true leaf stage judgment module;

The cotton five-true-leaf stage judging module is used to determine whether cotton has entered the five-true-leaf stage according to the distribution of the number of side stems in the cotton plant sub-image after the seedlings are settled: for the cotton plant sub-image containing the main stem of the plant, the side stem The intersection of the stem and the main stem is used as a node. When the number of sub-images of 2 or more nodes is detected to account for more than 50% of the number of sub-images of cotton plants after seedling setting, it is judged that the cotton field has entered the five true leaf stage.

Generally speaking, compared with the prior art, the above technical scheme conceived by the present invention can obtain the following beneficial effects due to the use of image processing method to analyze the photos of cotton fields, thereby judging whether the cotton fields have entered the five true leaf stage:

(1) Instead of manually judging the five true leaf stages of cotton, it saves manpower;

(2) Through the method of image processing, the state of the cotton field can be monitored in real time, and whether the cotton field has entered the five true leaf stage can be reported at any time, which is beneficial to agricultural meteorological observation;

(3) Analyzing images through precise segmentation methods, and judging the five true leaf stages through statistical data, which is more accurate and reliable than manual judgment;

(4) By rationally optimizing the image processing parameters and selecting a suitable image processing algorithm, taking into account the cotton field image processing speed and processing effect.

Description of drawings

Fig. 1 is the automatic detection method flowchart of cotton five true leaf stages provided by the present invention;

Fig. 2 is the front view of the cotton field;

Figure 3 is a front view of a cotton field where the growth of a single row of plants can be clearly observed;

Figure 4 is the result of cotton plant sub-image processing after seedling acquisition: Figure 4(a) is the original image of the plant sub-image, Figure 4(b) is the result of contrast stretching, and Figure 4(c) is the result of color segmentation. 4(d) is the Mean shift segmentation result graph, and Figure 4(e) is the comprehensive segmentation result graph;

Fig. 5 is the lower view of the cotton field;

Fig. 6 is the result map after green segmentation of Fig. 5;

Fig. 7 is the change trend chart of the cotton field lower view coverage;

Figure 8 is the result of detecting the main stem of cotton plants: Figure 8(a) is an example of a sub-image of a cotton plant, Figure 8(b) is a binary sub-image of the initial edge of the plant, and Figure 8(c) is the binary value of the initial skeleton of the plant Subgraph, Figure 8(d) is the result of the main stem detection;

Figure 9 is the result of detecting cotton plant side stems: Figure 9(a) is the left side of the cotton plant sub-image, Figure 9(b) is the right side of the cotton plant sub-image, and Figure 9(c) is the detection result of the side stem picture;

Fig. 10 is a node detection diagram of a cotton plant sub-image.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other.

The automatic detection method of the five true leaf stages of cotton provided by the invention comprises the following steps:

(1) Obtain the sub-image of the cotton plant after seedling setting.

Collect the front view of the cotton field after the seedlings have been settled. The height from the ground is 0.35 meters, the focal length of the lens is 14 mm, the horizontal shooting direction is east, the angle with the horizon is 0 degrees, and the resolution of the camera is not less than 4 million pixels. Taking each day as a detection period, the two cameras took w cotton images (w=13) in each detection period. Every day is a detection period, which is beneficial to identify the main key growth period of cotton. Since the object of our image processing is a single cotton plant, it is first necessary to cut the front view of the cotton field after seedling setting to obtain a front view of the cotton field that can clearly observe the growth status of a single row of plants. Since the single-row plants are distributed in the lower 1/3 of the whole picture in the sequence picture of the front view of the cotton field after seedling setting, the cut size of the image is preferably "the width of the front view of the cotton field" × (1/3 x "height of front view image"). After cutting, obtain the front view of the single row of plants in the cotton field after seedling setting.

Using a search box suitable for the image size of a single cotton plant, the image is split into sub-images with a certain split step. Preferably, the search box is equal to the height of the front view of the single row of plants, the width of the search box is 1/4 to 1/2 of its height, and the split step is 1/2 to 5/6 of the width. After observing the image of a single plant, it is found that the aspect ratio of the plant is generally 3:1. Since the size of the sub-image affects the integrity and accuracy of detection, the size of the sub-image can be more preferably the size of the search box and The height of the front view of the single row of plants is equal, and the width of the search box is 1/3 of its height. The split order affects the search program design, and the split order is preferably from left to right; the split step affects the time and accuracy of a single plant search, the larger the split step, the faster the split, the lower the accuracy, and the step size The smaller the splitting speed, the slower the splitting speed and the higher the accuracy, the splitting step is preferably 1/2 to 5/6 of the width of the search box, in order to ensure that each plant in a single row of plants is detected and to avoid repeated detection of the same plant Plants, more preferably 4/5 of the width of the search box.

In order to carry out subsequent detection more accurately, the method of contrast stretching can be used first to highlight the plants. For the stretched image: use different methods such as color segmentation or crop image segmentation to obtain a single plant segmentation result map. Cotton plant leaves and the upper part of the main stem are green, and the lower part of the main stem is reddish brown. Based on the color features of a single cotton plant, preferably, the sub-image is segmented into green and reddish brown using the relationship between the three color channels in the RGB (red/green/blue) color space. When performing green segmentation on an image, R, G/R, and G/B values are used as feature indicators. If the relationship between the three channel values of pixels in the image satisfies R≤200, G/R≥0.9 and G/B≥2.5 at the same time , then the green leaves can be basically segmented out; when reddish-brown image is segmented, R, R/G, B/G values are used as feature indicators, if the relationship between the three channel values of pixels in the image satisfies R>100 at the same time, If R/G≥2.8 and B/G≥0.1, the reddish-brown branches can be basically segmented, and the segmentation results of a single plant can be obtained by combining the green and reddish-brown segmentation results. The crop image segmentation method is used to filter the background noise of the image in the field natural environment by using spatial information. The crop segmentation method based on Meanshift is the preferred crop segmentation method to obtain the crop image segmentation result. Combine the results of crop image segmentation and color segmentation, that is, on the area segmentation result map of the crop segmentation method based on Mean shift, retain the area occupied by the detected pixels in the color segmentation result map to obtain the cotton after planting Plant subimage.

Due to the complexity of the image background in the natural environment of the field, when using the relationship between the three color channels in the RGB color space for color segmentation, the spatial information is not utilized, resulting in noise and pores in the segmentation results. Therefore, the crop image segmentation method based on Mean shift is used to segment the sub-image. Finally, combining the results of Mean shift and color segmentation, that is, combining space and color information, can segment plants more accurately and obtain a clearer sub-image of cotton plants.

The time for setting seedlings is determined as follows: collect the lower view images of the cotton field under the same conditions every day, preferably, a camera 5 meters above the ground is used, the focal length of the lens is 14 mm, the horizontal shooting direction is east, and the angle with the horizon is 60 degrees downward , the camera resolution is not less than 4 million pixels. Utilize segmentation method to carry out green segmentation to described image, count the proportion of green pixel in described image and be green image coverage; The green image coverage of the lower view of cotton field every day is compared with the green image coverage of cotton field lower view of the previous day, when When the coverage of the green image decreases, it is the seedling setting time. Described segmentation method can adopt methods such as environmental adaptive segmentation method, hypergreen operator segmentation method, crop image segmentation method based on Mean shift, Fisher linear discriminant method (referring to [1] Lei F.Tian.Environmentally adaptive segmentation algorithm for outdoor imagesegmentation.Computers and electronics in agriculture,1998,21:153～168); [2]D.M.Woebbecke,G.E.Meyer,K.Von Bargen,D.A.Mortensen.Color Indices for weed identification under various soil,residue,and lighting conditions of the Transactions ASAE,1995,38(1):259～269); [3] Zheng L, Zhang J, Wang Q. Mean-shift-based color segmentation of images containing green vegetation. Computers and Electronics in Agriculture, 2009, 65:93- 98); [4] Zheng L, Shi D, Zhang J. Segmentation of green vegetation of crop canopy images based on mean shift and Fisher linear discriminant. Pattern Recognition Letters, 2010, 31 (9): 920～925.).

(2) Detect the main stem of the cotton plant.

For all the cotton plant sub-images obtained in step (1), the edge detection algorithm is used to detect the initial edge of the plant to obtain the edge binary sub-image; the skeleton detection algorithm is used to extract the initial plant skeleton to obtain the skeleton binary sub-image. For the edge detection algorithm, image detection operators that can be used include Sobel operator, Roberts operator, LoG operator, and Canny operator, etc., and the Canny operator with strong edge integrity is preferred for edge detection. The skeleton detection algorithm is preferably a thinning operation in the morphological image processing method.

Chain code detection and straight line detection are performed on the edge binary subgraph and the skeleton binary subgraph, and the vertical plant edge and plant skeleton of the cotton plant are obtained. The edge binary submap and the skeleton binary submap are superimposed, and the inner side of the vertical plant edge containing the vertical plant skeleton is used as the main stem of the cotton plant to obtain the subimage of the cotton plant including the main stem of the plant. The straight line detection preferably adopts the Hough transform algorithm.

The main stem of the cotton plant has obvious straight-line edges, and the growth direction of the main stem is basically vertical. Therefore, the method of detecting the vertical line can be used to judge whether there is a main stem in the plant edge extraction sub-image. Extract the binary subgraph of the edge of the plant operator and the binary subgraph of the edge of the plant skeleton line, respectively first use the chain code to extract the straight line, and then use the Hough transform to detect the straight line result of the chain code extraction line, and finally the operator edge The detection straight line result of extracting the binary subgraph is merged with the detection straight line result of the skeleton line edge extraction binary subgraph, and the final main stem detection result can be obtained.

(3) Detection of lateral stems of cotton plants.

The cotton plant sub-image obtained in step (2) including the main stem of the plant is divided into two sides according to the position of the main stem: the left side of the main stem is the left side of the image, and the right side of the main stem is the right side of the image. Color segmentation and straight line detection are performed on the images on both sides to obtain the primary lateral stems in the images on both sides.

The principle and steps of color segmentation are as follows: when a cotton plant grows to the five-true leaf stage, the color of the side stem of a single plant will turn reddish-brown. According to the color characteristics of the side stem in the images on both sides, first use the three-channel pixel in the image to The relationship between the values, color segmentation of the sub-image of a single plant based on the main stem: using R, R/G, B/G values as feature indicators, if the relationship between the pixel three-channel values in the image satisfies R> 100, R/G≥2.8 and B/G≥0.1, the reddish-brown side stems can be basically separated.

Line detection is carried out on the images on both sides of the sub-image of a single plant based on the main stem after the color segmentation is completed. Preferably, a line detection method using Hough transform is used.

Among the primary lateral stems of the images on both sides, the straight line forming an acute angle with the upper side of the main stem was considered as the lateral stem.

(4) Determine the five true leaf stage of cotton.

For the sub-image of the cotton plant containing the main stem of the plant obtained in step (2), the intersection point of the side stem and the main stem is used as the node. When two or more nodes are detected, the number of sub-images accounts for all cotton plants When the number of sub-images is more than 50%, it is considered that the cotton field has entered the five true leaf stage, otherwise the next day's detection will be carried out.

For the convenience of program design, the number of nodes is used to represent the number of side stems. The nodes on a single plant are the intersection points of the main stem and side stems. It intersects with the main stem, and the point of intersection is determined as a node. Judging by experience, the image generally cannot collect all nodes, so when the number of sub-images with 2 or more nodes is detected to account for more than 50% of all sub-images, it is judged that the cotton field has entered the five-true leaf stage, otherwise it is judged that the cotton has not entered the five-true stage leaf stage.

According to the automatic detection method for the five true leaf stages of cotton, an automatic detection system for the five true leaf stages of cotton is provided, which is characterized in that it includes a cotton plant sub-image acquisition module, a cotton main stem detection module, a cotton side stem detection module, and a cotton side stem detection module. Five true leaf stage judgment module;

The cotton plant sub-image acquisition module is used to collect the front view of a single row of cotton plants in a cotton field after seedling setting, split them into individual cotton plant sub-images, and process the sub-images into cotton plant sub-images after seedling setting to transfer to cotton main stem detection module;

The cotton main stem detection module is used to extract the edge of the cotton plant and its main skeleton, and the inner side of the vertical plant edge containing the vertical plant skeleton is used as the main stem of the cotton plant to obtain a sub-image of the cotton plant including the main stem of the plant, and The sub-image is delivered to the cotton lateral stem detection module;

The cotton side stem detection module is used to divide the cotton plant sub-image including the main stem of the plant into two sides according to the position of the main stem: the left side of the main stem is the left side of the image, and the right side of the main stem is the right side of the image; For the initial lateral stems in the images on both sides, the straight line forming an acute angle with the upper side of the main stem is used as the lateral stems, and the detection results are passed to the cotton five-true leaf stage judgment module;

The cotton five-true-leaf stage judging module is used to determine whether cotton has entered the five-true-leaf stage according to the distribution of the number of side stems in the cotton plant sub-image after the seedlings are settled: for the cotton plant sub-image containing the main stem of the plant, the side stem The intersection of the stem and the main stem is used as a node. When the number of sub-images of 2 or more nodes is detected to account for more than 50% of the number of sub-images of cotton plants after seedling setting, it is judged that the cotton has entered the five true leaf stage, otherwise it is judged that the cotton has not entered Five true leaves stage.

The following are examples:

Use the method provided by the invention to judge whether the cotton in Fig. 2 enters the five true leaf stage:

(1) Obtain the sub-image of the cotton plant after seedling setting.

Collect the front view of the cotton field after the seedlings are settled, the camera is 0.35 meters above the ground, the focal length is 14 mm, the horizontal shooting direction is east, the angle with the horizon is 0 degrees, and the resolution is 4 million pixels. The embodiment uses each day as a detection period, The camera takes w cotton images (w=13) in each detection period, and the image size is 3648×2736. Figure 2 is an example of the front view of the cotton field. First, cut the front view of the cotton field after seedling setting to obtain the front view of the single row of plants in the cotton field after seedling setting, which can clearly observe the growth status of the single row of plants. The image size is 3648×912, as shown in Figure 3.

Using a search box of 912 pixels × 304 pixels and a split step of 240 pixels, in order from left to right, the image is split into sub-images of 912 pixels × 304 pixels that only contain a single plant, as shown in Figure 4 (a) shown.

For each sub-image, the plants are highlighted using the method of contrast stretching, as shown in Fig. 4(b). Color segmentation and crop segmentation operations are performed on the contrast-stretched sub-images, respectively. Color segmentation: R, G/R, and G/B values are used as feature indicators. If the relationship between the three channel values of pixels in the image satisfies R≤200, G/R≥0.9 and G/B≥2.5 at the same time, it can basically Segment the green leaves; when segmenting the reddish-brown image, use the R, R/G, and B/G values as feature indicators. If the relationship between the three channel values of the pixel in the image satisfies R>100 at the same time, R/G≥ 2.8 and B/G≥0.1, the reddish-brown branches can be basically segmented out, and the color segmentation results can be obtained by combining the green and reddish-brown segmentation results, as shown in Figure 4(c). Crop segmentation: The crop segmentation algorithm based on Mean shift is used to segment the image into crops, and the result is shown in Figure 4(d). Combine the results of crop image segmentation and color segmentation, that is, on the area segmentation result map of the crop segmentation method based on Mean shift, retain the area occupied by the detected pixels in the color segmentation result map to obtain the cotton after planting The plant sub-image is shown in Figure 4(e).

The seedling setting time is determined according to the following method: Under the same conditions, the cotton field is collected under the same conditions every day, the camera is 5 meters above the ground, the focal length is 14 mm, the horizontal shooting direction is east, the angle with the horizon is 60 degrees downward, and the resolution is 4 million pixels. In the embodiment, each day is regarded as a detection period, and the camera takes w cotton images (w=13) in each detection period, and the image size is 3648×2736, as shown in Figure 5, which is an example of the bottom view of the cotton field. Utilize Fisher's linear discriminant method to carry out green segmentation to described image, as shown in Figure 6 is the green segmentation result map, the green image coverage in the described image of statistics; Cotton field lower view green image coverage of every day after cotton field three true leaf stages Compared with the green image coverage of the lower view of the cotton field the day before, when the coverage of the green image decreases, it is the time for setting seedlings. Figure 7 shows the change trend of the coverage of the lower view of the cotton field. The decrease of coverage at day time indicates that the cotton farmers have completed the cotton field seedling operation on this day, that is, the cotton field transverse front-view image sequence can be used to detect the growth status of a single plant from the 10th day.

(2) Detection of the main stem of cotton plants

For all the sub-images of cotton plants obtained in step (1), as shown in Figure 8(a) is an example of a sub-image of cotton plants, use the Canny operator to detect the initial edge of the plant, and obtain the binary sub-image of the initial edge of the plant, as shown in As shown in Figure 8(b); the initial skeleton of the plant is extracted using the thinning operation in the morphological image processing method, and the binary subgraph of the initial skeleton of the plant is obtained, as shown in Figure 8(c).

The binary subgraph of the initial edge of the plant and the binary subgraph of the initial skeleton of the plant are firstly extracted by chain code, and then the result of the straight line extracted by the chain code is detected by Hough transform to obtain the vertical plant edge and A binary subgraph of the plant skeleton. The plant edge binary submap and the plant skeleton binary submap are superimposed, and the inner side of the vertical plant edge containing the vertical plant skeleton is used as the main stem of the cotton plant to obtain the subimage of the cotton plant including the main stem of the plant, as shown in Figure 8(d ) shown.

(3) Detection of lateral stems of cotton plants

The cotton plant sub-image obtained in step (2) including the main stem of the plant is divided into two sides according to the position of the main stem: the left side of the main stem is the left side of the image, as shown in Figure 9(a), and the right side of the main stem is is the right side of the image, as shown in Figure 9(b). Color-segment the images on both sides: color-segment the images on both sides to obtain the reddish-brown side stems that are basically segmented, and use R, R/G, and B/G values as feature indicators. If the relationship among them satisfies R>100, R/G≥2.8 and B/G≥0.1 at the same time, then the reddish-brown lateral stem can be basically segmented to obtain the basically segmented reddish-brown lateral stem, and then Hough is performed on the color segmentation result map Straight line detection gets the straight lines in the images on both sides.

In the images on both sides, the straight line forming an acute angle with the upper side of the main stem is considered as a side stem, and the side stem detection result can be obtained by combining the images on both sides, as shown in Figure 9(c).

(4) Judging the five true leaf stage of cotton

The node on a single plant is the intersection of the main stem and the side stem. When there is no intersection between the main stem and the side stem as determined by the color segmentation and line detection results, the side stem is extended until it intersects with the main stem, and the intersection point is determined to be the node. Judging by experience, the image generally cannot collect all nodes. For the sub-image of cotton plants containing the main stem of the plant obtained in step (2), when the number of sub-images with 2 or more nodes is detected accounts for all the sub-images of cotton plants after seedling setting. When the number is more than 50%, it is considered that the cotton field has entered the five true leaf stage. That is, the sub-image when two or more lateral stems are detected, as shown in Figure 10.

It is easy for those skilled in the art to understand that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, All should be included within the protection scope of the present invention.

Claims (8)

1. one to grow cotton five leaf period detection methods, it is characterized in that, comprise the following steps:

(1) cotton plants subimage after acquisition final singling: the positive elevational views of the single plant in cotton field after collection final singling; With the search box of applicable cotton plants individual plant image size, described image is split into subimage; Color segmentation and crop Iamge Segmentation are carried out to described subimage, on the region segmentation result figure of crop dividing method, the region occupied by pixel detected in retaining color segmentation result figure, cotton plants subimage after acquisition final singling;

(2) cotton plants stem is detected: for all cotton plants subimages obtained in step (1), detect plant just edge, extract plant just skeleton, to plant just edge and plant just skeleton carry out Chain Code Detection and straight-line detection obtains the vertical plant edge of cotton plants and plant skeleton, using inside the vertical plant edge comprising vertical plant skeleton as cotton plants stem, obtain and comprise the cotton plants subimage of plant stem;

(3) cotton plants side stem is detected: by the cotton plants subimage comprising plant stem obtained in step (2), according to stem position, be divided into both sides: stem is on the left of image with a left side, stem is on the right side of image with the right side; Both sides image is carried out respectively the first side stem that color segmentation and straight-line detection obtain in the image of both sides, will wherein with stem on the upside of acutangulate just side stem as side stem;

(4) cotton five leaf period is judged: for the cotton plants subimage comprising plant stem obtained in step (2), using the intersection point of wherein side stem and stem as node, when cotton plants number of sub-images more than 50% after the number of sub-images of more than 2 or 2 nodes accounts for all final singlings being detected, judge that cotton field enters five leaf periods, otherwise judge that cotton field does not enter five leaf periods.

2. cotton five leaf period detection method as claimed in claim 1, it is characterized in that, described step (1) final seedling time judges according to following steps: every day gathers view image under cotton field under the same conditions, utilize dividing method to carry out green segmentation to described image, add up described image Green pixel proportion and be green image coverage; View green image coverage under every day cotton field is compared with view green image coverage under the previous day cotton field, is final seedling time when green image coverage reduces.

3. cotton five leaf period detection method as claimed in claim 2, it is characterized in that, the described dividing method carrying out green segmentation and use, can adopt environment self-adaption dividing method, super green operator dividing method, the crop image partition method based on Mean shift, Fisher linear discriminant method.

4. cotton five leaf period detection method as claimed in claim 1, it is characterized in that, the positive elevational views of described single plant is through the method process of contrast stretching.

5. cotton five leaf period detection method as claimed in claim 4, it is characterized in that, described search box is equal with the height of the positive elevational views of described single plant, and the width of described search box is 1/4 to 1/2 of its height, and described fractionation step-length is 1/2 to 5/6 of described search box width.

6. cotton five leaf period detection method as claimed in claim 1, it is characterized in that, described step (2) detects plant just edge algorithms and extraction plant just skeleton, and adoptable image detective operators has Sobel operator, Roberts operator, LoG operator and Canny operator, preferred Canny operator.

7. cotton five leaf period detection method as claimed in claim 1, is characterized in that, described step (2) and step (3) cathetus detect and can adopt Hough transform.

8. one to grow cotton five leaf period detection systems, it is characterized in that, comprise cotton plants subimage acquisition module, cotton stem detection module, cotton side stem detection module and cotton five leaf period judge module;

Described cotton plants subimage acquisition module, for gathering the positive elevational views of the single plant in cotton field after final singling, split into cotton plants individual plant subimage, and after described subimage is processed into final singling, cotton plants subimage passes to cotton stem detection module;

Described cotton stem detection module, for extract cotton plants edge and main framing, using inside the vertical plant edge comprising vertical plant skeleton as cotton plants stem, obtain and comprise the cotton plants subimage of plant stem, and described subimage is passed to cotton side stem detection module;

Described cotton side stem detection module, for comprising the cotton plants subimage of plant stem, according to stem position, be divided into both sides: stem is on the left of image with a left side, stem is on the right side of image with the right side; Obtain the first side stem in both sides images, will wherein with stem on the upside of acutangulate straight line as side stem, and testing result is passed to cotton five leaf period judge module;

Described cotton five leaf period judge module, for the distribution situation according to stem number in side in cotton plants subimage after final singling, judge whether cotton enters five leaf periods: for the cotton plants subimage comprising plant stem, using the intersection point of wherein side stem and stem as node, when cotton plants number of sub-images more than 50% after the number of sub-images of more than 2 or 2 nodes accounts for all final singlings being detected, judge that cotton enters five leaf periods, otherwise judge that cotton does not enter five leaf periods.