CN-121767866-B - Termite indicator identification method and system based on optical and infrared image data

CN121767866BCN 121767866 BCN121767866 BCN 121767866BCN-121767866-B

Abstract

The invention provides a termite indicator recognition method and a termite indicator recognition system based on optical and infrared image data, which relate to the technical field of image recognition, and the termite indicator recognition method comprises the steps of obtaining a gray optical image of a current dyke, carrying out spectrum analysis on each pixel in the gray optical image to obtain an optimal direction and an optimal micro-periodic texture response of the pixel, screening out pixels with qualified optimal micro-periodic texture response as candidate pixels, and forming a plurality of indicator direction chains by the candidate pixels based on a path expansion rule set in the optimal direction; and setting an extension pixel judgment rule, connecting corresponding indicator chains based on the extension pixel judgment rule, merging the indicator direction chains based on an abutting relation between the indicator direction chains, dividing an optical image of the dam into an indicator region, an indicator undetermined region and other regions based on the size of the merged region, acquiring infrared image data, and analyzing to further identify the indicator undetermined region.

Inventors

XU HUAN
GAO YONGYONG
LIU LONG

Assignees

延安大学

Dates

Publication Date: 20260505
Application Date: 20260303

Claims (9)

1. A termite indicator identification method based on optical and infrared image data is characterized by comprising the following specific steps: step 1, acquiring a current gray level image of a dyke, constructing a sliding window with a pixel as a center for each pixel in the gray level image, gradually rotating the sliding window and performing spectrum analysis on an internal image of the sliding window to obtain the optimal direction and the optimal micro-periodic texture response of the pixel; Step 2, screening out pixels with qualified response of the optimal micro-periodic texture as candidate pixels, and forming a plurality of indicator direction chains by the candidate pixels based on a path expansion rule set in the optimal direction; Step 3, for each end point pixel of each indicator direction chain, determining extension pixels in the neighborhood of each end point pixel based on extension pixel judgment rules to update the end point pixel, and iterating until the preset iteration times are reached or intersection points exist between other indicator direction chains, so as to obtain updated indicator chains; And 4, merging the indicator direction chains based on the adjacent relation between the indicator direction chains, dividing the optical image of the dam into an indicator area, an indicator undetermined area and other areas based on the size of the merged indicator direction chains, acquiring the infrared images of the dam at the current moment and the previous moment, performing one-to-one mapping with the optical image of the dam, performing thermal inertia analysis on the indicator undetermined area based on the infrared images to obtain thermal inertia characteristics, setting an indicator judging rule based on the thermal inertia characteristics, and taking the indicator undetermined area as the indicator area if the indicator undetermined area meets the indicator judging rule, otherwise, taking the indicator undetermined area as the other areas.
2. The termite indicator identification method based on optical and infrared image data according to claim 1, wherein the logic for determining the optimal direction and the optimal micro-periodic texture response of any pixel in a gray level image is that a sliding window taking the pixel as a center is constructed, the rotating sliding window is rotated clockwise according to a preset rotation amplitude, the spectral energy concentration of the pixel in the sliding window during each rotation is recorded until one rotation is completed, the maximum spectral energy concentration is used as the optimal micro-periodic texture response of the pixel, and the corresponding rotation angle is used as the optimal direction of the pixel; The calculation logic of the spectrum energy concentration degree is that Fourier transformation is carried out on all pixels in a sliding window to obtain spectrum data of the pixels in the sliding window, power spectrum density under each frequency is calculated, frequency corresponding to the maximum value of the power spectrum density is obtained to obtain a main frequency, a main frequency bandwidth is determined based on a relative peak power method, spectrum energy in a frequency range of the main frequency bandwidth is calculated and is called main frequency energy, and the ratio of the main frequency energy to the total spectrum energy is calculated to obtain the spectrum energy concentration degree of the pixels in the sliding window.
3. The termite indicator identification method based on the optical and infrared image data according to claim 2, wherein the logic for determining the bandwidth of the main frequency based on the relative peak power method is to calculate the product of the power spectral density of the main frequency and the threshold ratio, called the relative peak power, obtain all frequencies with the power spectral density equal to the relative peak power as candidate frequencies, select the minimum value of the candidate frequencies among the candidate frequencies larger than the main frequency, called the bandwidth maximum value, select the maximum value of the candidate frequencies among the candidate frequencies smaller than the main frequency, called the bandwidth minimum value, and the interval formed by the bandwidth minimum value and the bandwidth maximum value is called the main frequency bandwidth.
4. The termite indicator identification method based on optical and infrared image data according to claim 1, wherein the logic of screening out the pixels with qualified optimal micro-periodic texture response is to preset an optimal micro-periodic texture response threshold, and pixels with optimal micro-periodic texture response greater than the optimal micro-periodic texture response threshold are called candidate pixels.
5. The termite indicator recognition method based on optical and infrared image data according to claim 1 is characterized in that a path expansion rule is that for any candidate pixel, a connection line of which the center points to the center of each neighborhood pixel is obtained, the connection line is called a neighborhood connection line, a ray which takes the center of the candidate pixel as a starting point and extends along the optimal direction of the candidate pixel is called a main direction ray, an included angle between each neighborhood connection line and the main direction ray is calculated and is used as a departure angle between a corresponding neighborhood pixel and the candidate pixel, a neighborhood pixel with the minimum departure angle is extracted and is used as a same chain pixel of the candidate pixel, if the same chain pixel of one candidate pixel is still the candidate pixel, the two candidate pixels are mutually the same chain pixel, the two candidate pixels form a chain set, all candidate pixels are traversed, if an intersection exists between any two chain sets, the two chain sets are combined in sequence until the intersection does not exist in all the chain sets, and the pixels in each chain set form an indicator direction chain.
6. The termite indicator recognition method based on optical and infrared image data according to claim 5, wherein the rule of the extension pixel judgment is that, for each end pixel of the indicator direction chain, the candidate pixel which is the same as the end pixel of the chain is directed to the direction of the end pixel as the extension direction, the ray extending along the extension direction with the center of the end pixel as the starting point is called the extension direction ray, the included angle between each neighborhood connecting line and the extension direction ray is calculated as the extension angle of the corresponding neighborhood pixel and the end pixel, and the neighborhood pixel with the minimum extension angle is extracted as the extension pixel of the end pixel; The logic for updating the extension pixel into a new endpoint pixel and performing iteration is that the extension pixel is updated into the endpoint pixel, and the endpoint pixel pairs before and after updating are regarded as candidate pixel pairs of the same chain pixel, and the iteration is performed.
7. The termite indicator recognition method based on the optical and infrared image data according to claim 1, wherein the logic of dividing the optical image of the dam into an indicator region, an indicator pending region and other regions is that in any two indicator direction chains, if any one pixel of one indicator direction chain is in an adjacent relation with any one pixel of the other indicator direction chain, all pixels in the two indicator direction chains are combined into one indicator direction chain, iterating until all indicator direction chains are traversed, after iterating is completed, each indicator direction chain is called as a candidate region, a candidate region area threshold is preset, for any candidate region, if the area of the candidate region is larger than the candidate region area threshold, the candidate region is divided into the indicator region, if the area of the candidate region is not larger than the candidate region area threshold, the candidate region is divided into the indicator pending region, and in the optical image of the dam, if the area of the non-candidate region is divided into other regions.
8. The termite indicator recognition method based on optical and infrared image data according to claim 1, wherein the logic for setting the indicator judgment rule is that for each indicator pending area, the temperature change rate of each pixel in the indicator pending area at two moments is calculated based on the infrared images at the current moment and the previous moment, and the average value of the temperature change rates of all pixels in the indicator pending area is calculated and is called pending temperature inertia; For all indicator areas, calculating the temperature change rate of each pixel in all indicator areas at two moments, and calculating the average value of the temperature change rates of all pixels in all indicator areas, which is called standard temperature inertia; And calculating a relative error between the inertia of the undetermined temperature and the inertia of the standard temperature, namely an inertia difference value, wherein the indicator judgment rule is that the inertia difference value of the undetermined area of the indicator is smaller than a preset inertia difference value threshold value.
9. A termite indicator identification system based on optical and infrared image data is characterized in that the system is used for realizing the termite indicator identification method based on optical and infrared image data according to any one of claims 1-8, and specifically comprises the following steps: The optimal analysis module is used for acquiring a current gray level image of the dam, constructing a sliding window taking the pixel as a center for each pixel in the gray level image, gradually rotating the sliding window and carrying out spectrum analysis on an internal image of the sliding window so as to obtain the optimal direction and the optimal micro-periodic texture response of the pixel; The chain construction module is used for selecting the pixels with qualified response of the optimal micro-periodic texture as candidate pixels and forming a plurality of indicator direction chains by the candidate pixels based on the path expansion rules set in the optimal direction; The chain extension module is used for determining extension pixels in the neighborhood of each end point pixel based on extension pixel judgment rules for each end point pixel of each indicator direction chain so as to update the end point pixel, and iterating until the preset iteration times are reached or intersection points exist between other indicator direction chains so as to obtain updated indicator chains; The merging identification module is used for merging the indicator direction chains based on the adjacent relation between the indicator direction chains, dividing the optical image of the dykes into an indicator area, an indicator undetermined area and other areas based on the size of the merged indicator direction chains, acquiring the infrared images of the dykes at the current moment and the previous moment, carrying out one-to-one mapping with the optical image of the dykes, carrying out thermal inertia analysis on the indicator undetermined area based on the infrared images to obtain thermal inertia characteristics, setting an indicator judging rule based on the thermal inertia characteristics, and taking the indicator undetermined area as the indicator area if the indicator undetermined area meets the indicator judging rule, otherwise taking the indicator undetermined area as the other areas.

Description

Termite indicator identification method and system based on optical and infrared image data Technical Field The invention relates to the technical field of image recognition, in particular to a termite indicator recognition method and system based on optical and infrared image data. Background Aiming at the existing termite inspection work, the termite inspection work is mostly dependent on manual visual or single optical image to identify the indicators such as termite mud lines, mud covers, branch flying holes and the like. The method is basically based on color difference, shape characteristic or simple texture difference to judge, is extremely easy to be interfered by soil block deposition, scouring trace, shadow, water stain and vegetation coverage under a complex natural background, and has high misjudgment rate and poor stability. In recent years, unmanned aerial vehicle optical and infrared data are introduced for auxiliary identification, but the prior art mostly adopts a feature superposition type fusion idea, namely, optical texture features and infrared thermal abnormal features are respectively extracted, and fusion judgment is carried out through a threshold value or a model. However, termite mud is not typically a significant goal, it does not necessarily have a distinct color difference optically, nor is it a heat source or heat sink in the infrared, but simply by repeated particle packing to form a dense layer with a microcycled structure and alter the local heat conduction path. Therefore, the existing method cannot identify from the termite mud formation mechanism, so that the identification reliability is insufficient, and the termite indicator is difficult to effectively distinguish from natural deposition, a wet zone or surface disturbance traces. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The present invention is directed to a termite indicator recognition method and system based on optical and infrared image data, so as to solve the above-mentioned problems in the prior art. In order to achieve the above purpose, the present invention provides the following technical solutions: A termite indicator identification method based on optical and infrared image data comprises the following specific steps: step 1, acquiring a current gray level image of a dyke, constructing a sliding window with a pixel as a center for each pixel in the gray level image, gradually rotating the sliding window and performing spectrum analysis on an internal image of the sliding window to obtain the optimal direction and the optimal micro-periodic texture response of the pixel; Step 2, screening out pixels with qualified response of the optimal micro-periodic texture as candidate pixels, and forming a plurality of indicator direction chains by the candidate pixels based on a path expansion rule set in the optimal direction; Step 3, for each end point pixel of each indicator direction chain, determining extension pixels in the neighborhood of each end point pixel based on extension pixel judgment rules to update the end point pixel, and iterating until the preset iteration times are reached or intersection points exist between other indicator direction chains, so as to obtain updated indicator chains; And 4, merging the indicator direction chains based on the adjacent relation between the indicator direction chains, dividing the optical image of the dam into an indicator area, an indicator undetermined area and other areas based on the size of the merged indicator direction chains, acquiring the infrared images of the dam at the current moment and the previous moment, performing one-to-one mapping with the optical image of the dam, performing thermal inertia analysis on the indicator undetermined area based on the infrared images to obtain thermal inertia characteristics, setting an indicator judging rule based on the thermal inertia characteristics, and taking the indicator undetermined area as the indicator area if the indicator undetermined area meets the indicator judging rule, otherwise, taking the indicator undetermined area as the other areas. Further, for any pixel in the gray level image, the logic for determining the optimal direction and the optimal micro-periodic texture response of the pixel is that a sliding window taking the pixel as the center is constructed, the rotating sliding window rotates clockwise according to the preset rotation amplitude, the spectrum energy concentration degree of the pixel in the sliding window is recorded when the sliding window rotates each time until the sliding window rotates one circle, the maximum spectrum energy concentration degree is used as the optimal micro-periodic texture response of the pixel, and the corr