CN116046698A

CN116046698A - Potato detection method, device, equipment, storage medium and program product

Info

Publication number: CN116046698A
Application number: CN202211714990.6A
Authority: CN
Inventors: 王文秀; 张凡; 马倩云; 孙剑锋; 刘亚琼; 牟建楼; 王颉
Original assignee: Hebei Agricultural University
Current assignee: Hebei Agricultural University
Priority date: 2022-12-29
Filing date: 2022-12-29
Publication date: 2023-05-02

Abstract

The invention provides a potato detection method, a device, equipment, a storage medium and a program product, which relate to the technical field of agriculture and comprise the following steps: carrying out hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato; extracting color characteristic information and texture characteristic information in the image of the target potato to obtain image characteristic information of the target potato; extracting characteristic parameters of volatile gas of the target potatoes to obtain odor characteristic information of the target potatoes; inputting the multi-source characteristic information of the target potato into a trained detection model, and outputting a dry rot prediction diagnosis result of the target potato, wherein the multi-source characteristic information comprises the hyperspectral characteristic information, the image characteristic information and the smell characteristic information.

Description

Potato detection method, device, equipment, storage medium and program product

Technical Field

The invention relates to the technical field of agriculture, in particular to a potato detection method, a potato detection device, potato detection equipment, a storage medium and a program product.

Background

With the rapid development of the potato industry, the number and period of potato storage has further increased and prolonged in order to accommodate market demands. However, during storage, potatoes are extremely susceptible to pathogen infection, various storage period diseases are initiated, and the quality of the potatoes is seriously affected. Wherein, dry rot is one of the most common fungus diseases in the storage period, and is caused by infection of fusarium, fungus spores are infected to the surface of the potato, and continuously reproduce after the period of submerged cultivation, and the disease is changed from recessive to dominant, so that the external and internal diseases of the potato are finally caused. The perennial morbidity of the potato is up to 10% -30%, the highest morbidity can be up to more than 60%, and the potato is extremely easy to infect surrounding normal individuals, so that the commodity value and the eating quality of the potato are seriously affected, and the potato becomes a serious disease for limiting the development of the potato industry. Therefore, the rapid diagnosis and detection of the potatoes under the stress of the dry rot are realized, especially the early identification of the peristalsis period is significant for preventing and treating diseases.

The traditional method for identifying and detecting the dry rot is to identify the fusarium morphologically, but the method is time-consuming and labor-consuming, and the culture characteristics of the fusarium can be obviously changed due to the various and complex fusarium classification systems and the change of environmental conditions such as temperature, humidity, illumination, pH and the like. In recent years, development of immunological technology, PCR technology and the like provides a new method for detecting pathogenic bacteria, but the methods still have a plurality of defects, such as that the detection performance of an enzyme-linked immunosorbent assay (ELISA) is greatly dependent on the performance of an antibody, a detection kit cannot be reused, and the detection cost is high; the fluorescent quantitative PCR method requires high expertise and is not suitable for large-scale rapid real-time monitoring.

Therefore, how to better detect the dry rot of the potato has become a problem to be solved in the industry.

Disclosure of Invention

The invention provides a potato detection method, a device, equipment, a storage medium and a program product, which are used for solving the problem that how to better detect the dry rot of potatoes in the prior art is urgent to solve in the industry.

The invention provides a potato detection method, which comprises the following steps:

carrying out hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato;

extracting color characteristic information and texture characteristic information in the image of the target potato to obtain image characteristic information of the target potato;

extracting characteristic parameters of volatile gas of the target potatoes to obtain odor characteristic information of the target potatoes;

inputting the multi-source characteristic information of the target potato into a trained detection model, and outputting a dry rot prediction diagnosis result of the target potato, wherein the multi-source characteristic information comprises the hyperspectral characteristic information, the image characteristic information and the smell characteristic information.

According to the potato detection method provided by the invention, hyperspectral image correction and characteristic spectrum extraction are carried out on the hyperspectral image of the target potato to obtain hyperspectral characteristic information of the target potato, and the method comprises the following steps:

correcting the hyperspectral image based on Beer-Lambert reflection law and an illuminance-reflection model to obtain a corrected hyperspectral image;

acquiring first spectrum data of a region of interest in the corrected hyperspectral image, and performing filtering and first derivative processing on the first spectrum data to obtain second spectrum data;

performing two-dimensional correlation spectrum analysis on the sensitive wave band in the second spectrum data to obtain and analyze a corresponding two-dimensional synchronous spectrum and an autocorrelation spectrum to obtain hyperspectral characteristic information of the target potato; wherein the sensitive wave band is a wave band with obvious difference of potatoes with different disease degrees.

According to the potato detection method provided by the invention, the color characteristic information and the texture characteristic information in the image of the target potato are extracted to obtain the image characteristic information of the target potato, and the method comprises the following steps:

carrying out color feature extraction on the image of the target potato by utilizing the color moment and the histogram to obtain color feature information;

extracting texture features of the image of the target potato by using a gray level co-occurrence matrix to obtain texture feature information;

and obtaining the image characteristic information of the target potato according to the color characteristic information and the texture characteristic information.

According to the potato detection method provided by the invention, the characteristic parameters of the volatile gas of the target potato are extracted to obtain the odor characteristic information of the target potato, and the method comprises the following steps:

acquiring sensor array data corresponding to the volatile gas, constructing an instantaneous linear mixing matrix and a mixing signal analysis matrix according to the sensor array data, and establishing a mixing signal analysis model;

after carrying out de-averaging, whitening and moving average treatment on the sensor array data, constructing a signal-to-noise ratio objective function;

and carrying out gradient solving and extreme point analysis on the signal-to-noise ratio objective function to obtain a separation matrix in the signal-to-noise ratio objective function so as to obtain smell characteristic information of the target potato.

According to the potato detection method provided by the invention, the multisource characteristic information of the target potato is input into the trained detection model, and before the step of outputting the dry rot prediction diagnosis result of the target potato, the method further comprises the following steps:

acquiring hyperspectral characteristic sample information, image characteristic sample information and air quality characteristic sample information of a sample potato;

taking hyperspectral characteristic sample information, image characteristic sample information and air quality characteristic sample information of the sample potatoes and corresponding dry rot disease degree labels as a training sample to obtain a plurality of training samples;

and training the preset convolutional neural network by utilizing the plurality of training samples.

According to the potato detection method provided by the invention,

training a preset convolutional neural network by using the plurality of training samples, wherein the training comprises the following steps:

for any training sample, inputting the training sample into the preset convolutional neural network, and outputting a dry rot prediction diagnosis result corresponding to the training sample;

calculating a loss value according to the dry rot prediction diagnosis result and the dry rot disease degree label in the training sample by using a preset loss function;

and if the loss value is smaller than a preset threshold value, completing the training of the preset convolutional neural network.

The invention also provides a potato detection device, which comprises:

the first extraction module is used for carrying out hyperspectral image correction and characteristic spectrum extraction on the hyperspectral image of the target potato to obtain hyperspectral characteristic information of the target potato;

the second extraction module is used for extracting color characteristic information and texture characteristic information in the image of the target potato to obtain the image characteristic information of the target potato;

the third extraction module is used for extracting characteristic parameters of the volatile gas of the target potatoes to obtain smell characteristic information of the target potatoes;

the detection module is used for inputting the multi-source characteristic information of the target potato into a trained detection model and outputting a dry rot prediction diagnosis result of the target potato, wherein the multi-source characteristic information comprises the hyperspectral characteristic information, the image characteristic information and the smell characteristic information.

According to the potato detection device provided by the invention, the device is further used for:

The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the potato detection method as described in any one of the above when executing the program.

The invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which when executed by a processor implements a potato detection method as described in any of the above.

The invention also provides a computer program product comprising a computer program which when executed by a processor implements a potato detection method as described in any one of the above.

According to the potato detection method, the device, the equipment, the storage medium and the program product, provided by the invention, the rapid detection of the dry rot potato can be realized in a limited way by combining the multi-source characteristic information consisting of the hyperspectral characteristic information, the image characteristic information and the smell characteristic information of the target potato and the deep convolution neural network, and meanwhile, the problems of locality, hysteresis, destructiveness, indirection and the like of the traditional disease detection method are solved.

Drawings

In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic flow chart of a potato detection method provided in an embodiment of the present application;

fig. 2 is a schematic structural diagram of a potato hyperspectral image capturing device in an embodiment of the present application;

FIG. 3 is a schematic diagram of a potato odor characteristic information acquisition process in an embodiment of the present application;

fig. 4 is a schematic structural diagram of a potato detection apparatus according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Fig. 1 is a schematic flow chart of a potato detection method provided in an embodiment of the present application, as shown in fig. 1, including:

step 110, performing hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato;

fig. 2 is a schematic structural diagram of a potato hyperspectral image capturing device in an embodiment of the present application, as shown in fig. 2, where the device includes a translation stage, a potato sample may be placed on the translation stage, and a light source, a lens, a spectrometer and a CCD camera may be sequentially disposed above the translation stage.

Specifically, the hyperspectral image described in the embodiments of the present application may specifically be a hyperspectral image that is set in advance by the hyperspectral image processing software including exposure time, start-stop position, moving speed, light source intensity, and the like. Black and white image correction is then performed. And finally, placing the sample on a translation stage, adjusting an initial scanning position, and controlling a camera to acquire line scanning hyperspectral images one by one to obtain the hyperspectral image of the target potato.

Because the surface of the potato has a certain radian, the hyperspectral image can be further corrected and adjusted, and then the characteristic extraction is carried out, so that hyperspectral characteristic information of the target potato is obtained.

Step 120, extracting color feature information and texture feature information in the image of the target potato to obtain image feature information of the target potato;

specifically, in the embodiment of the application, color feature extraction is performed by using the color moment and the histogram, texture feature information is extracted by using the gray level co-occurrence matrix, and finally image feature information of the target potato is obtained.

Step 130, extracting characteristic parameters of volatile gas of the target potato to obtain smell characteristic information of the target potato;

in the embodiment of the application, the gas sensor sensitive to the specific odor component is screened to form a special sensor array with specific recognition on the dry rot specific volatile. According to the type property presumption of the volatile gas of the dry rot potatoes, combining with the early investigation, selecting a surface resistance control type gas sensor-SnO 2 type gas sensor, selecting different types of sensors to form an array according to the response characteristics of the sensors, designing a corresponding signal conditioning circuit to realize analog-to-digital conversion, acquiring odor sensing signals of samples with different disease degrees, and analyzing the target potatoes in the mode to obtain odor characteristic information of the target potatoes.

And 140, inputting the multi-source characteristic information of the target potato into a trained detection model, and outputting a dry rot prediction diagnosis result of the target potato, wherein the multi-source characteristic information comprises the hyperspectral characteristic information, the image characteristic information and the smell characteristic information.

When potatoes are infested with fungi, the external characteristics, internal components and tissue structure of the potatoes are changed, and the optical characteristics of the potatoes are changed correspondingly. The hyperspectral imaging is a novel technology for fusing images and spectrums, has the characteristics of high resolution, more wave numbers and 'spectrum unification', can simultaneously acquire the image information of each wavelength and the spectrum information of each pixel of a detected object, and provides possibility for early diagnosis and detection of potato dry rot. On the other hand, fusarium on potato growth metabolism can cause degradation of its own compounds, generate characteristic volatile odor, and can be used as an 'odor fingerprint mark' of a sample for real-time diagnosis of dry rot. The odor sensing technology provides a new thought and a new method for early identification of the dry rot infection of the potatoes, and the hyperspectral imaging technology can obtain information reflecting the degree of the dry rot disease of the potatoes from different angles in complementary advantages, and the fusion of the two can effectively improve the early nondestructive rapid diagnosis effect of the dry rot disease of the potatoes.

Therefore, in the embodiment of the application, hyperspectral characteristic information, image characteristic information and smell characteristic information are taken as input, one-dimensional data are converted into two-dimensional data, so that the capability of deep mining of the characteristic information of the convolutional neural network model is enhanced, and finally, a CNN model is built by taking the artificially marked disease label as an output variable. Because the convolution kernel of each layer of convolution is randomly generated in the forward transmission process of the network, the characteristic output of the final layer of convolution layer is different, the whole network model is required to be iterated continuously, the random quantity in the training process is optimized according to the loss between the predicted value and the label value of the network, the model fitting is finally achieved, and the optimal model parameters are saved. And adopting the cross entropy loss function as a measurement index of model updating.

In the embodiment of the application, based on the acquired map and smell information, a Fuzzy Clustering (FCM) algorithm, a Dynamic Time Warping (DTW) algorithm and the like are combined to determine the time sequence key points of the dry rot from the peristalsis period to the morbidity period. And further fusing the extracted characteristics through a full-connection layer and Softmax, and constructing an early diagnosis prediction model of the dry rot potatoes with different disease degrees under the controllable conditions of artificial inoculation by taking the disease degree as an output layer.

Finally, the model is verified by taking the potato with the dry rot in a natural state as a test object, and the potato dry rot prediction diagnosis model with strong applicability is constructed by optimizing and improving an ROI selection method, a characteristic wavelength extraction method, a depth image characteristic extraction method, an odor sensing characteristic parameter, a cluster analysis algorithm and the like.

In the embodiment of the application, the rapid detection of the dry rot potato can be realized in a limited way by combining the multi-source characteristic information consisting of the hyperspectral characteristic information, the image characteristic information and the smell characteristic information of the target potato and the deep convolutional neural network, and meanwhile, the detection method also solves the problems of locality, hysteresis, destructiveness, indirection and the like of the traditional disease detection method.

Optionally, performing hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato, including:

performing two-dimensional correlation spectrum analysis on the sensitive wave band in the second spectrum data to obtain and analyze a corresponding two-dimensional synchronous spectrum and an autocorrelation spectrum to obtain hyperspectral characteristic information of the target potato; wherein the sensitive wave band is a wave band corresponding to potatoes with different disease degrees.

Specifically, in the embodiment of the application, based on Beer-Lambert reflection law and illuminance-reflection model, the system compares the advantages and disadvantages of brightness correction methods such as elliptical brightness transformation, low-pass filtering, B-sample surface fitting brightness correction model, homomorphic filtering and the like, and determines a surface self-adaptive brightness correction method suitable for potatoes.

Spectral data is extracted from the ROI selected from each wavelength image, and noise signals such as baseline drift and stray light are reduced by utilizing pretreatment methods such as wavelet denoising and direct signal correction.

Then, taking the infection time as an external perturbation, selecting the spectrum of samples with different inoculation days as a representative spectrum, and comparing with a differential treatment and envelope removal method, selecting sensitive wave bands with obvious differences between the corresponding spectrums of potatoes with different disease degrees. And finally, respectively carrying out two-dimensional correlation spectrum analysis on the sensitive wave bands to acquire and analyze two-dimensional synchronous spectrums and autocorrelation spectrums of the sensitive wave bands, thereby defining characteristic information closely related to the dry rot.

In the embodiment of the application, the hyperspectral image is corrected, so that the accuracy of data can be effectively ensured, two-dimensional correlation spectrum analysis is respectively carried out on sensitive wave bands, and the two-dimensional synchronous spectrum and the autocorrelation spectrum of the hyperspectral image can be acquired and analyzed, so that the characteristic information closely related to the dry rot is clear.

Optionally, extracting color feature information and texture feature information in the image of the target potato to obtain image feature information of the target potato, including:

Specifically, in the embodiment of the application, images of 647nm, 550nm and 460nm are selected as RGB images, and first-order moments and second-order moments of each component of RGB are extracted to represent color distribution of the images. The color information of the RGB color space is mapped into the HSV color space, and the color histogram of each component of the HSV is obtained, so that the defect of nonuniform RGB color space is overcome. And extracting texture characteristic information by using the gray level co-occurrence matrix. The gray level co-occurrence matrix of four different angles (0 ℃, 45 ℃, 90 ℃ and 135 ℃) is obtained, and contrast, correlation, entropy, uniformity and energy are used as feature scalar for representing the gray level co-occurrence matrix, so that the texture condition of the image is more intuitively and carefully described.

In the embodiment of the application, the image feature extraction can be effectively carried out on the image of the target potato, and the subsequent construction of multi-source data for potato inspection is facilitated.

Optionally, extracting characteristic parameters of volatile gas of the target potato to obtain odor characteristic information of the target potato, including:

acquiring sensor array data corresponding to the volatile gas, constructing an analysis matrix according to the sensor array data, and establishing a mixed signal analysis model;

Specifically, in the embodiment of the application, taking potato samples with different disease degrees as objects, weighing 5g of a sample to be measured, placing the sample into a 20mL headspace sample injection bottle, and sealing. After incubation at 90℃for 30min, the sample was introduced at a needle temperature of 85℃and a sample volume of 100. Mu.L. GC-IMS conditions: MXT-5 column (15 m. Times.0.53 mm,4 μm); the incubation rotating speed is 500r/min, the column temperature is 60 ℃, the carrier gas is N2, the IMS temperature is 45 ℃, and the operation analysis time is 20min. Carrier gas flow gradient setting: the initial flow rate is 2mL/min, and the flow rate is kept for 2min; linearly rising to 10mL/min within 2-10 min; and (5) linearly rising to 100mL/min within 10-20 min, and stopping.

Based on the analysis, the gas sensor sensitive to the specific odor component is screened to form a special sensor array with specific recognition to the dry rot specific volatile. According to the type property presumption of the volatile gas of the dry rot potato, combining with the early investigation, the surface resistance control type gas sensor-SnO 2 type gas sensor is selected, different types of sensors are selected to form an array according to the response characteristics of the sensors, and a corresponding signal conditioning circuit is designed to realize analog-to-digital conversion, so that the odor sensing signals of samples with different disease degrees are obtained.

Fig. 3 is a schematic diagram of a process for acquiring smell characteristic information of potatoes in the embodiment of the application, as shown in fig. 3, a potato sample may be placed in a closed chamber, volatile gas of the potato sample may enter a sensor array chamber through a pipeline, a special sensor array having specific identification on dry rot-specific volatile matters in the sensor array may generate a weak voltage signal according to the volatile gas of the potato sample, and then the weak voltage signal passes through a signal conditioning circuit and a microprocessor to finally obtain smell characteristic information of potatoes.

The obtained volatile gas composition of the potatoes has the characteristics of multiple types and multi-source mixing, and the gas sensor array has the problems of cross sensitivity and the like in the response process, so that the accuracy of the odor sensing signals is low, and the accuracy and the reliability of detection are affected. Therefore, the multi-source mixed signal of the dry rot potato is analyzed based on the principle of mutual informatization, the analysis matrix of the multi-source mixed signal is obtained, the high-precision separation of multi-sensing data is realized, and the volume fraction of the gas component parameter is quantitatively analyzed. And then, carrying out range normalization processing on sensor array data after smoothing and denoising, extracting characteristic values and characteristic parameters such as a maximum response value, a maximum response first derivative value, a maximum recovery first derivative value, a minimum second derivative response value and the like from signals such as a filtered signal, a differential signal and the like by using a time domain characteristic parameter extraction method, and obtaining an optimal characteristic parameter set through correlation analysis to obtain smell characteristic information of the target potato.

In the embodiment of the application, the odor characteristic information of the potatoes can be effectively obtained by analyzing the odor information of the potatoes, so that the subsequent construction of multi-source data is facilitated for potato detection.

Optionally, the multi-source characteristic information of the target potato is input into a trained detection model, and before the step of outputting the dry rot prediction diagnosis result of the target potato, the method further comprises:

Specifically, the preset convolutional neural network described in the embodiment of the present application may specifically be composed of a convolutional layer, a pooling layer, and a fully-connected layer, and the neural network proposed in the embodiment of the present invention includes a single convolutional layer composed of 20 convolutional filters 9*9. The output of the convolution layer enters the pooling layer after passing through the ReLU function. The pooling layer adopts the average pooling of 2 x 2 submatrices. The neural network classifier includes a single hidden layer and an output layer. The hidden layer has 100 nodes and adopts a ReLU activation function. Since there are 10 classes, the output layer takes 10 nodes and the activation function is Softmax.

Table 1 shows parameters of the neural network model, specifically:

TABLE 1 model parameters

/>

Based on the acquired atlas and smell information, a Fuzzy Clustering (FCM) algorithm, a Dynamic Time Warping (DTW) algorithm and the like are combined to determine the time sequence key points from the peristalsis period to the morbidity period. And further fusing the extracted characteristics through a full-connection layer and Softmax, and constructing an early diagnosis prediction model of the dry rot potatoes with different disease degrees under the controllable conditions of artificial inoculation by taking the disease degree as an output layer.

In the embodiment of the application, the detection of the dry rot of the potatoes can be effectively realized through a trained detection model. The detection method also solves the problems of locality, hysteresis, destructiveness, indirection and the like of the traditional disease detection method.

The potato detection apparatus provided by the invention is described below, and the potato detection apparatus described below and the potato detection method described above can be referred to correspondingly.

Fig. 4 is a schematic structural diagram of a potato detection apparatus provided in an embodiment of the present application, as shown in fig. 4, including:

the first extraction module 410 is configured to perform hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato, so as to obtain hyperspectral characteristic information of the target potato;

the second extraction module 420 is configured to extract color feature information and texture feature information in the image of the target potato, so as to obtain image feature information of the target potato;

the third extraction module 430 is configured to extract characteristic parameters of the volatile gas of the target potato, so as to obtain odor characteristic information of the target potato;

the detection module 440 is configured to input multi-source characteristic information of the target potato into a trained detection model, and output a predicted diagnosis result of the dry rot of the target potato, where the multi-source characteristic information includes the hyperspectral characteristic information, the image characteristic information, and the odor characteristic information.

Fig. 5 is a schematic structural diagram of an electronic device according to the present invention, and as shown in fig. 5, the electronic device may include: processor 510, communication interface (Communications Interface) 520, memory 530, and communication bus 540, wherein processor 510, communication interface 520, memory 530 complete communication with each other through communication bus 540. Processor 510 may invoke logic instructions in memory 530 to perform a potato detection method comprising: carrying out hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato;

Further, the logic instructions in the memory 530 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

In another aspect, the present invention also provides a computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing the potato detection method provided by the methods described above, the method comprising: carrying out hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato;

In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the potato detection method provided by the above methods, the method comprising: carrying out hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato;

The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A potato detection method, comprising:

2. The potato detection method of claim 1, wherein performing hyperspectral image correction and characteristic spectrum extraction on a hyperspectral image of a target potato to obtain hyperspectral characteristic information of the target potato comprises:

acquiring first spectrum data of a region of interest in the corrected hyperspectral image, and performing filtering processing on the first spectrum data to obtain second spectrum data;

3. The potato detection method of claim 1, wherein extracting color feature information and texture feature information in the image of the target potato to obtain the image feature information of the target potato comprises:

4. The potato detection method of claim 1, wherein extracting characteristic parameters of volatile gas of the target potato to obtain odor characteristic information of the target potato comprises:

5. The potato detection method of claim 1, wherein the step of inputting the multi-source characteristic information of the target potato into a trained detection model and outputting the dry rot prediction diagnosis result of the target potato is preceded by the step of:

6. The potato detection method of claim 5, wherein training a preset convolutional neural network using the plurality of training samples, comprises:

and if the loss value is smaller than a preset threshold value, completing training of the preset convolutional neural network to obtain a trained detection model.

7. A potato detection apparatus, comprising:

8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the potato detection method of any one of claims 1 to 6 when executing the program.

9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the potato detection method of any of claims 1 to 6.

10. A computer program product comprising a computer program which, when executed by a processor, implements a potato detection method as claimed in any one of claims 1 to 6.