A method for detecting the flowering process and flowering quality of peach trees

Abstract

The present invention discloses a method for detecting the flowering process and flowering quality of peach trees, which relates to the field of fruit tree growth status monitoring and perception. The method comprises: detection window image cropping, initial flower opening time and initial flower withering time detection, detection window flowering process statistics, and flowering quality assessment. The method overcomes the limitation of traditional monitoring and investigation methods that consume too much manpower, and effectively realizes automatic quantitative monitoring of peach flower growth.

Description

Technical Field

[0001] The present invention relates to the field of fruit tree growth condition monitoring, and in particular to a method for detecting the flowering process and flowering quality of peach trees. Background Art

[0002] Peach trees are prolific flowering trees, and their flowering process and quality vary depending on a variety of factors, both internal to the plant and external to the environment. Adverse weather conditions during the flowering period, such as high temperatures, low temperatures, frost, and rain, can lead to rapid flower and fruit drop and low fruit set, necessitating timely agronomic measures to preserve flowers and fruit. Under favorable conditions, peach trees produce abundant and high-quality flowers, requiring appropriate flower and fruit thinning to ensure stable yield and quality. As global warming intensifies, average temperatures are rising in most regions, and the frequency and intensity of extreme heat waves are increasing. In many areas, temperatures fluctuate dramatically during the spring flowering period for fruit trees, impacting flowering and fruit set.

[0003] By monitoring the flowering process of peach trees, we can assess the quality of flower growth and provide a reference for understanding fruit set and early fruit drop, guiding subsequent management measures such as fruit thinning and fruit preservation. Traditionally, flowering information relies primarily on manual on-site observation to record flowering, flower drop, and fruit drop, lacking automated quantitative methods for monitoring peach blossom growth. Currently, some orchards use surveillance cameras to record orchard images. Various object detection and instance segmentation models can be used to identify flowering trees and extract flowering characteristics. However, the density of peach blossoms, overlapping and obscuring each other, as they gradually bloom, makes deep learning models less effective in detecting flower opening and initial flower drop. Summary of the Invention

[0004] The object of the present invention is to provide a method for detecting the flowering process and flowering quality of peach trees, extracting the initial opening time and initial withering time of flowers in multiple detection windows in a monitoring image, counting the duration of opening of each window, and evaluating the growth status of peach flowers according to the ratio of the number of windows with different opening durations and the corresponding flowering quality weights, so as to realize automatic acquisition of information such as the flowering process and quality status by using monitoring images.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for detecting the flowering process and flowering quality of a peach tree comprises the following steps:

[0007] Detection window image cropping;

[0008] Detection of initial flower opening time and initial flower withering time;

[0009] Statistics of flowering process in detection window;

[0010] Flowering quality assessment;

[0011] In some embodiments, the detection window image cropping includes: setting a number of detection windows distributed on the branches of the peach tree, covering the branches without overlapping, and having the same length and width in the fruit tree growth image collected at a fixed point and a fixed observation angle, and using the detection windows to crop the image to obtain a sub-image set W to be processed.

[0012] In some embodiments, the detection of the initial opening time and initial withering time of the flower includes: using a deep learning model to detect whether there is a flower target in the window image, and the time corresponding to the first appearance of the flower target is the initial opening time d1 of the flower; using the time series safflower index calculated by the window time series image to detect the initial withering time of the flower, and detecting the safflower index after d1. The time when the value rises to the maximum value and then begins to decline is the initial withering time d2.

[0013] In some embodiments, the red flower index is the ratio of the difference between the mean values ​​of the red channel and the green channel of the detection window image to the mean value of the green channel. The time series red flower index can be subjected to SG filtering to reduce noise.

[0014] In some embodiments, the flowering process statistics of the detection window include extracting the initial opening time d1 and the initial withering time d2 of each window, calculating the complete opening duration (d2-d1+1), and counting the number of windows detected with flowering process and the proportion of windows with different opening durations.

[0015] In some embodiments, the flowering quality assessment is performed by multiplying the proportions of windows with different opening durations by the flowering quality weights of the corresponding windows and then accumulating the results.

[0016] In some embodiments, the window blooming quality weight includes: the weight of less than normal duration, the weight of normal duration, and the weight of more than normal duration. Taking the normal duration as a reference, the weight of less than normal duration is smaller than the weight of normal duration, and the weight of more than normal duration is larger than the weight of normal duration.

[0017] The beneficial effects of the method disclosed in this application for detecting the flowering process and flowering quality of peach trees include but are not limited to:

[0018] The final result is the key time information of initial opening and initial withering reflecting the local and overall flowering process of peach trees; according to the proportion of flowering in the detection window, the phenological information of the initial flowering period, the peak flowering period, the withering period and other phenological information of traditional observations can be obtained, and the flowering quality of the whole tree can be evaluated according to the proportion and weight of the windows with different durations of opening. By setting a reasonable detection window, the adverse effects of some image background objects on growth monitoring are reduced; the growth appearance process after the flower blooms is described by the red flower index time series data of the present invention, while the errors caused by image quality and image processing algorithms are reduced, and the time information related to the pollination and fertilization of peach flowers is reasonably reflected, so as to realize the flower quality evaluation; compared with the traditional manual observation and judgment, the present invention realizes the expression of quantitative indicators and accurately reflects the details of the key process after flowering; compared with the single reliance on deep learning methods, it avoids the problem of poor recognition accuracy of flower withering when flowers are clustered. The present invention has better adaptability in digital information orchards, and can achieve relatively stable recognition effects for flower growth image data taken regularly by automated image acquisition equipment under different environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of monitoring the peach blossom growth process according to one embodiment of the present invention;

[0020] Figure 2 A schematic diagram of the setting and distribution of peach blossom growth image detection windows according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the principle of extracting the initial opening and initial withering time of peach blossoms according to an embodiment of the present invention;

[0022] Figure 4 This is a diagram showing the detection results of the peach tree flowering process according to one embodiment of the present invention;

[0023] Figure 5 This is a diagram showing the peach tree flowering quality detection results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0025] On the contrary, this application covers any alternatives, modifications, equivalents, and solutions made within the spirit and scope of this application as defined by the claims. Furthermore, to facilitate a better understanding of this application, certain specific details are described in detail below in the detailed description of this application. Those skilled in the art will be able to fully understand this application without these details.

[0026] The following is a detailed description of a method for monitoring the growth process and quality of peach blossoms according to an embodiment of the present invention. It should be noted that the following embodiments are merely used to explain the present invention and do not constitute a limitation of the present invention.

[0027] like Figure 1 As shown, a method for monitoring the growth process and quality of peach blossoms comprises the following steps:

[0028] Step S01, detecting window image cropping,

[0029] The detection window is set according to the distribution of branches during the budding period. The number of windows is not less than 30, and the area covers the main branches without overlapping. The length and width of each window are the same. The monitoring image of the i-th day is cropped using n windows to obtain the sub-image set to be processed.

[0030] Step S02: Detect the initial opening time and initial falling time of the flowers.

[0031] The initial opening time of the flower is the time when the petals first open in the window time series image, using the date array Expression, using the flower target deep learning model to detect the sub-image set W to be processed i , if there is an image of window k When the first flowering is detected, the date of the window in the array is updated according to the corresponding date of the i-th day

[0032] The initial falling time is the time when the petals begin to fall or wilt in the window time series image, using the date array The red flower index RFI is determined by calculating the image of each window. When the RFI increases to the maximum and then begins to decline, the corresponding date is the initial withering time. The red flower index is the ratio of the difference between the red channel mean R and the green channel mean G of the window image and the green channel mean. The image sequence of window k is According to the formula Calculate the time series RFI If the window image When a flower is detected to be open, the date corresponding to the first maximum value after flowering is extracted according to the time series RFI of the window as the initial flower withering time

[0033] Step S03, detecting the statistics of the window blooming process,

[0034] Extract the initial flowering time d1 and initial withering time d2 of each window, calculate the duration of complete opening (d2-d1+1), and count the number of windows with flowering detected and the proportion of windows with different opening durations;

[0035] Step S04, flowering quality assessment,

[0036] According to the accumulation of the ratio of different flowering duration windows and the product of the corresponding flowering quality weights, the flowering quality weights include: the weight of less than the normal duration, the weight of the normal duration, and the weight of more than the normal duration. Taking the normal duration as a reference, the weight of less than the normal duration is smaller, and the weight of more than the normal duration is larger.

[0037] The weight of less than normal duration, the weight of normal duration, and the weight of more than normal duration are based on the normal duration. The weight of less than normal duration is smaller, and the weight of more than normal duration is larger.

[0038] The following is an example of a specific embodiment:

[0039] The growth of fruit trees is monitored and collected at a fixed point with a fixed observation angle. In order to realize flowering detection and index calculation, the monitoring image data includes at least red, green and blue channel values.

[0040] Step S01, detection window image cropping, set 33 windows of 224×224 pixels on the non-flowering branches, the area covers the main branches and does not overlap, crop the monitoring image of the i-th day, the effect is shown in Figure 2 , get the sub-image set to be processed;

[0041] Step S02, flower initial opening time, initial withering time detection, according to Figure 3 The principle of extracting the initial opening and initial withering time of peach flowers is shown. The flower target deep learning model is used to detect the flowering targets in each window of the sub-image set to be processed, and the initial opening time of the flower is obtained. The initial withering time of the flower is obtained according to the corresponding date when the red flower index RFI increases to the maximum and then begins to decrease.

[0042] Step S03: Detect the statistics of the flowering process of the window, extract the initial flowering time d1 and the initial withering time d2 of the window with flowering process, and calculate the duration of complete opening (d2-d1+1). Figure 4 , the number of windows with flowering process and the proportion of windows with different opening time are counted. The effect is shown in Figure 5 ;

[0043] Step S04, flowering quality evaluation, referring to the relationship between the flowering time of the detection window and the fruit setting rate, the weight of the opening time less than the normal opening time is set to 0.1, the weight of the normal opening time is set to 0.6, and the weight of the opening time more than the normal opening time is set to 0.9. The proportion of the windows with different opening time is multiplied by the flowering quality weight of the corresponding window, and the flowering quality evaluation result is accumulated. Figure 5 .

Claims

1. A method for detecting the flowering process and flowering quality of peach trees, characterized in that: The following steps are involved: Detection window image cropping; Detection of initial flower opening time and initial flower withering time; Statistics of flowering process in detection window; Flowering quality assessment; The detection of the initial opening time and initial withering time of the flower comprises: using a deep learning model to detect the time corresponding to the first appearance of a flower target in the window image as the initial opening time of the flower, and determining the initial withering time according to the time when the maximum value of the red flower index in the image begins to decrease; The flowering quality evaluation includes first calculating the product of the ratio of windows with different opening time lengths and the flowering quality weight of the corresponding windows, and then accumulating the results; The red flower index is the ratio of the difference between the mean values ​​of the red channel and the green channel of the detection window image to the mean value of the green channel; The window flowering quality weights include: the weight of less than normal duration, the weight of normal duration, and the weight of more than normal duration. Taking the normal duration as a reference, the weight of less than normal duration is smaller than that of normal duration, and the weight of more than normal duration is larger than that of normal duration.

2. A method for detecting the flowering process and flowering quality of peach trees according to claim 1, characterized in that, The detection window image cropping is to use multiple windows to crop the image to obtain a set of sub-images to be processed.

3. A method for detecting the flowering process and flowering quality of peach trees according to claim 1, characterized in that, The flowering process statistics of the detection windows include statistics of the initial opening time, initial withering time, opening duration, and the ratio of the number of windows with different opening durations in each detection window.

4. A method for detecting the flowering process and flowering quality of peach trees according to claim 2, characterized in that, The detection window includes a plurality of detection windows which are arranged according to the distribution of branches, have non-overlapping areas covering the branches, and have the same length and width.

Images