Target recognition-based fruit picking method, system, device and storage medium

Abstract

The application provides a fruit picking method, system and device based on target recognition and a storage medium, and relates to the technical field of computer vision.The application collects image data of a fruit to be recognized, performs preprocessing, constructs a fruit maturity prediction model by using a convolutional neural network, inputs the preprocessed image data into the model, outputs a maturity index of the fruit, further extracts key features in the fruit image, calculates the difference between the maturity stage and the fruit image features, generates a difference index, corrects the maturity index of the fruit, sets a maturity threshold, and judges whether the fruit has reached a picking standard.The application realizes accurate judgment of fruit maturity and automatic picking by combining image processing, feature data analysis and a machine learning model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, in particular to a fruit picking method, system, device and storage medium based on target recognition. BACKGROUND

[0002] Currently, the research and application of fruit picking methods are rapidly developing towards automation and intelligence. Traditional manual picking not only has high labor intensity and low efficiency, but also easily causes damage to fruits, affecting the yield. With the progress of computer vision, deep learning, and robotics, automated fruit picking systems based on target recognition have gradually become the focus of research. These systems can accurately identify the location of fruits through sensors and algorithms, analyze the maturity of fruits in real time, and complete precise picking through mechanical arms and other devices. Although the technology still faces challenges such as fruit types and environmental adaptability, it has shown great potential in improving efficiency, reducing costs, and promoting agricultural modernization.

[0003] The fruit picking method based on target recognition uses computer vision, deep learning, and other technologies to achieve automatic identification and precise picking of fruits, which can greatly improve agricultural production efficiency, reduce labor costs, and improve the accuracy of picking and the quality of fruits. This technology is crucial for the intelligent and automated development of modern agriculture, especially in addressing labor shortages, improving productivity, and achieving sustainable agriculture. With the continuous progress of artificial intelligence and robotics, this technology will be widely applied in various aspects of agricultural production, promoting the modernization of the agricultural industry and helping to solve the increasingly serious global food security problem.

[0004] The existing fruit picking methods mainly include manual picking and mechanized picking. Manual picking is the most traditional method, which has high labor intensity and low efficiency, especially in large-scale orchards, with high costs and cannot guarantee the picking quality. Mechanized picking uses various picking machines, robots, or mechanical arms to complete the work, which can improve the picking efficiency, but still has some problems. If the identification and precision are not accurate, the existing mechanical picking equipment often relies on visual sensors such as cameras and image processing algorithms for fruit identification and positioning. In the face of complex fruit tree structures and environmental changes, the recognition accuracy is often not high, and the fruits may be damaged;

[0005] Secondly, the existing picking equipment is usually suitable for specific varieties or specific fruit trees, and has certain limitations in fruit maturity, shape, size, etc., resulting in poor adaptability to different varieties and growth environments. Climate and environmental factors have a great impact on the stability and working efficiency of the picking equipment, especially in harsh weather or complex terrain, the equipment may not function properly.

[0006] Therefore, it is necessary to provide a fruit picking method, system, device and storage medium based on target recognition to solve the problem.

[0007] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0008] The purpose of the present application is to provide a fruit picking method, system, device and storage medium based on target recognition to solve the problems raised in the background.

[0009] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0010] A fruit picking method based on target recognition, the specific steps comprising:

[0011] Step 1: Collecting the image data of the fruit to be identified and pre-processing the image data of the fruit to be identified, the image preprocessing including denoising and enhancing contrast;

[0012] Step 2: Constructing a fruit maturity prediction model using a convolutional neural network, the fruit maturity prediction model including an input layer, a hidden layer and an output layer, the input layer being used to receive the pre-processed image data, the hidden layer being used to process the image data, and the output layer being used to output the maturity index of the fruit, and inputting the image data of the fruit to be identified into the prediction model to obtain the maturity index of the fruit to be identified;

[0013] Step 3: Extracting the feature data of the fruit to be identified in the image of the fruit to be identified, the feature data including roundness, contrast, energy and entropy, and obtaining the feature data of the fruit maturity stage, calculating the difference value between the maturity stage and the feature data of the image of the fruit to be identified, generating a difference index using the image feature difference value, and correcting the maturity index of the fruit to be identified based on the difference index to obtain a corrected maturity index of the fruit to be identified;

[0014] Step 4: Setting a maturity threshold, comparing the corrected maturity index of the fruit to be identified with the maturity threshold, and determining whether the target fruit can be picked.

[0015] Further, the image preprocessing includes denoising and enhancing contrast, and the method used is:

[0016] The denoising preprocessing of the image data reduces the noise points in the image by using a filtering technique, removes unnecessary noise while preserving image information, and ensures the accuracy of subsequent feature extraction. Secondly, the contrast enhancement enhances the contrast of local regions and better preserves image details, thereby improving the accuracy of the fruit recognition and picking model.

[0017] Further, a fruit maturity prediction model is constructed to obtain the maturity index of the to-be-identified fruit, and the method is as follows:

[0018] The fruit maturity prediction model adopts a convolutional neural network structure, including an input layer, a hidden layer and an output layer. The input layer is used to receive the preprocessed image data, the hidden layer is used to process the preprocessed image data, and the output layer is used to output the maturity index of the fruit .

[0019] The fruit maturity prediction model is trained, the historical fruit image data set is obtained, and the maturity index of the fruit is determined based on the expert score. The maximum fruit maturity index is selected as the training target, the mean square error function is selected as the loss function, the loss function value is calculated according to the output result and the real label, the gradient is calculated through the back propagation algorithm, and the weights and biases of the neural network are updated. Repeat the above operation until the model reaches the predetermined training number of rounds.

[0020] Further, the difference value between the maturity stage and the current to-be-identified fruit image feature data is calculated, and a difference index is generated using the image feature difference value. The formula is as follows:

[0021]

[0022]

[0023]

[0024]

[0025] wherein, , , respectively represent the difference values of the roundness, contrast, energy and entropy of the image feature data, , , is the roundness, contrast, energy and entropy of the collected fruit image feature data in the maturity stage, , , is the roundness, contrast, energy and entropy of the collected current to-be-identified fruit image feature data;

[0026] A difference degree index is generated by using the image feature difference value, and the formula is:

[0027]

[0028] Wherein, represents the difference degree index.

[0029] Further, the maturity index of the to-be-identified fruit is corrected based on the difference degree index to obtain a maturity index correction value of the to-be-identified fruit, and the formula is:

[0030]

[0031] Wherein, represents the maturity index correction value of the to-be-identified fruit, the maturity index of the fruit.

[0032] Further, the maturity index correction value of the to-be-identified fruit is compared with a maturity threshold value to determine whether the picking target fruit can be picked, and the logical formula is:

[0033]

[0034] Wherein, represents the established maturity threshold value, is a logical value for determining whether the picking target fruit can be picked, when , the maturity index correction value of the to-be-identified fruit is less than the maturity threshold value, indicating that the to-be-identified fruit is not mature and cannot be picked; when , the maturity index correction value of the to-be-identified fruit is greater than or equal to the maturity threshold value, indicating that the to-be-identified fruit is mature, and the picking operation is immediately performed.

[0035] The present application also provides a fruit picking system based on target recognition, which is used to execute the fruit picking method based on target recognition described above, and comprises:

[0036] A feature data acquisition module is configured to acquire to-be-identified fruit image data and pre-process the to-be-identified fruit image, wherein the image pre-processing includes denoising and contrast enhancement.

[0037] A data preprocessing module is configured to construct a fruit maturity prediction model by using a convolutional neural network, wherein the fruit maturity prediction model comprises an input layer, a hidden layer and an output layer, the input layer is configured to receive the pre-processed image data, the hidden layer is configured to process the image data, and the output layer is configured to output the maturity index of the fruit, and the to-be-identified fruit image data is input into the prediction model to obtain the maturity index of the to-be-identified fruit.

[0038] The picking model establishment module is used to extract feature data of the fruit to be identified from the image of the fruit to be identified. The feature data includes roundness, contrast, energy and entropy. It also obtains feature data of the fruit maturity stage, calculates the difference value between the maturity stage and the feature data of the image of the fruit to be identified, generates a difference index using the image feature difference value, and corrects the maturity index of the fruit to be identified based on the difference index to obtain the corrected maturity index value of the fruit to be identified.

[0039] The threshold determination module is used to set a maturity threshold, compare the maturity index correction value of the fruit to be identified with the maturity threshold, and determine whether the target fruit can be harvested.

[0040] This invention provides a computer device, comprising:

[0041] The electronic device includes a memory for storing a computer program and a processor for running the computer program to enable the electronic device to perform the above-described fruit-picking method based on target recognition.

[0042] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer, when executed by a processor, implements the above-described fruit picking method based on target recognition.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention, through the construction of a fruit recognition and harvesting model and more efficient sensors, can accurately identify fruits in different environments and complex structures, significantly reducing fruit damage rates and exhibiting greater adaptability. It is applicable to various varieties and shapes of fruits and can intelligently determine the optimal harvesting time based on fruit maturity. Based on a target recognition-based fruit harvesting method, the system can accurately analyze the fruit's color, plumpness, and gloss, and, combined with the fruit's firmness and texture characteristics, construct a maturity index model. This technology effectively determines whether fruits have reached the optimal harvesting time, thus avoiding the uncertainty of human experience-based judgment and improving harvesting accuracy and efficiency. Furthermore, through the application of convolutional neural networks, the system can handle complex orchard environments, achieving a high degree of automation, reducing manual labor, and lowering labor costs.

[0045] This invention, by calculating immature and overripe correction coefficients and combining them with standardized data processing methods, enables precise adjustment of fruit harvesting timing. By correcting the maturity index, fruit is harvested at its optimal ripeness, avoiding fruit waste caused by premature harvesting and losses due to overripe fruit rotting. This intelligent correction mechanism maximizes the economic value of the fruit, improving the overall output and economic benefits of the orchard. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the overall method flow of the present invention.

[0047] Figure 2 This is a schematic diagram of the system module flow of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0049] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0050] Example:

[0051] Please see Figure 1 The fruit picking method based on target recognition includes the following steps:

[0052] Step 1: Collect image data of the fruit to be identified and preprocess the image of the fruit to be identified. The image preprocessing includes noise reduction and contrast enhancement.

[0053] Step 2: Construct a fruit maturity prediction model using a convolutional neural network. The fruit maturity prediction model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive preprocessed image data, the hidden layer is used to process the image data, and the output layer is used to output the fruit maturity index. The image data of the fruit to be identified is input into the prediction model to obtain the maturity index of the fruit to be identified.

[0054] Step 3: Extract the feature data of the fruit to be identified from the image of the fruit to be identified. The feature data includes roundness, contrast, energy and entropy. Also, obtain the feature data of the fruit's ripening stage. Calculate the difference value between the ripening stage and the feature data of the image of the fruit to be identified. Use the image feature difference value to generate a difference index. Based on the difference index, correct the ripening index of the fruit to be identified to obtain the corrected ripening index value of the fruit to be identified.

[0055] Step 4: Set a maturity threshold, compare the maturity index correction value of the fruit to be identified with the maturity threshold, and determine whether the target fruit can be harvested.

[0056] It should be noted that denoising removes noise from the image using filtering techniques, ensuring the accuracy of subsequent feature extraction; contrast enhancement improves the details of local areas through adaptive methods, helping to improve the accuracy of the fruit recognition model; and scaling the image data to a uniform scale ensures that all data are on the same scale, thus providing high-quality input data for model training. These preprocessing steps work together to improve the accuracy of fruit recognition and ripeness prediction.

[0057] Therefore, image preprocessing includes denoising and contrast enhancement, based on the following methods:

[0058] Denoising preprocessing of image data involves using filtering techniques to reduce noise in the image. While preserving image information, unnecessary noise is removed to ensure the accuracy of subsequent feature extraction. Secondly, contrast enhancement is performed by using adaptive contrast enhancement to enhance the contrast of local areas, which can better preserve image details and thus improve the accuracy of the fruit recognition and picking model.

[0059] It is important to note that the process of constructing the fruit maturity prediction model is crucial, as it directly determines the accuracy and predictive effectiveness of the fruit maturity index. Employing a convolutional neural network structure enables the model to effectively process and extract features from images, thereby accurately reflecting the fruit's maturity state. By training on a historical fruit dataset based on expert scores, selecting mean squared error as the loss function, and utilizing backpropagation to optimize the network's weights and biases, the model can progressively approximate the true maturity index. This process ensures the model's high accuracy and reliability, providing a solid foundation for fruit identification and harvesting.

[0060] Therefore, it is necessary to construct a fruit maturity prediction model to obtain the maturity index of the fruit to be identified. The method used is as follows:

[0061] The fruit maturity prediction model employs a convolutional neural network structure, comprising an input layer, hidden layers, and an output layer. The input layer receives preprocessed image data, the hidden layers process the preprocessed image data, and the output layer outputs the fruit maturity index. Among them, the maturity index of the fruit to be identified The higher the value, the better. This index is a quantitative indicator for measuring the ripeness of fruit, and an increase in this value indicates that the fruit is gradually ripening and the ripeness is higher.

[0062] The fruit maturity prediction model is trained by acquiring a historical fruit image dataset and determining the fruit maturity index based on expert scores. The training objective is to maximize the fruit maturity index. The mean squared error function is selected as the loss function. The loss function value is calculated based on the output results and the true labels. The gradient is calculated through the backpropagation algorithm, and the weights and biases of the neural network are updated. The above operations are repeated until the model reaches the predetermined number of training rounds.

[0063] It is important to note that calculating the difference between the maturity stage and the image feature data of the fruit to be identified, and generating a difference index, is crucial for accurate prediction of fruit maturity. By analyzing the differences in image features such as roundness, contrast, energy, and entropy, subtle differences between different maturity stages can be better revealed, thus providing a reliable basis for subsequent maturity prediction. The calculation of the difference index combines these feature differences and processes them through mathematical formulas to derive a comprehensive index that accurately reflects changes in fruit maturity. This method not only improves the accuracy of predictions but also provides scientific support for the application of agricultural technologies such as automated harvesting.

[0064] Therefore, it is necessary to calculate the difference in image feature data between the ripening stage and the current fruit to be identified, and then generate a difference index using the image feature difference values. The formula used is as follows:

[0065]

[0066]

[0067]

[0068]

[0069] in, , , These represent the differences in roundness, contrast, energy, and entropy of the image feature data, respectively. , , The roundness, contrast, energy, and entropy of the collected fruit image feature data at the ripening stage are analyzed. , , The features of the fruit image to be identified include roundness, contrast, energy, and entropy.

[0070] The formula used to generate the difference index from image feature differences is as follows:

[0071]

[0072] in, This represents the difference index; in the above formula, , , , Increasing the value of will lead to A smaller value indicates a lower difference index, meaning greater differences between images and lower similarity. This suggests a significant difference between the maturity of the fruit to be identified and the maturity of a fruit at the ripe stage. Roundness describes how close an object's shape is to a circle. The roundness of a ripe fruit usually differs from that of the fruit to be identified; ripe fruit approaches a perfect circle, while unripe fruit is slightly irregular. This difference is addressed by squaring the difference values. To increase the impact of roundness differences, especially when the difference is large, the square operation is designed to ensure that large differences are not ignored. Entropy is an important indicator of image complexity, describing the randomness and information content of an image. Mature and immature fruits have different structural entropies; mature fruits may exhibit lower complexity and variation, i.e., lower entropy, while immature fruits may exhibit higher entropy. This can be addressed by using an exponential function. The formula assigns higher weights to entropy values ​​with larger differences because a large entropy difference may indicate a significant difference in image complexity, which is exponentially amplified. Contrast describes the difference in brightness regions of an image; ripe and unripe fruits exhibit significant differences in contrast, with ripe fruits often displaying more vivid colors and higher contrast. This is achieved by taking the cube root of the difference values. The formula moderately suppresses the impact of large contrast differences. This design prevents large contrast differences from excessively affecting the difference index, ensuring its consistency with the difference indices of other features. Energy is usually related to the smoothness or texture uniformity of an image. Mature fruits tend to have higher and more uniform energy, while unripe fruits may exhibit more local inhomogeneities. When the difference between the two is small, a logarithmic function is used. This design ensures that small differences do not cause drastic changes in the difference index, meaning it is less sensitive to energy changes in small differences, while amplifying larger differences, especially when energy differences are large, increasing their impact on the results. This design adapts to the importance of energy differences to image difference; the formula... It reflects the degree of difference between mature and immature fruits across multiple characteristics by weighted combination of various difference values. When the difference in a certain characteristic, such as roundness, entropy, contrast, or energy, is large, it will lead to... An increase in the value of the image, indicating a larger difference, suggests that the fruit to be identified is unripe; while when... A smaller value indicates a smaller difference in characteristics, meaning that the fruit to be identified is more similar to a mature fruit and has a higher degree of maturity.

[0073] It should be noted that correcting the maturity index of the fruit to be identified based on the difference index can more accurately reflect the similarity between the fruit to be identified and mature fruits. This correction process adjusts the maturity index by quantifying the differences in image features, making it more closely reflect the actual maturity state of the fruit. By introducing the difference index as a correction factor, errors caused by differences in image features can be effectively eliminated, thereby improving the accuracy of maturity assessment and ensuring that the maturity index of the fruit is more realistic and reliable. This method is of great significance for automated fruit maturity identification systems, and can provide more efficient support for fruit sorting and quality control in practical applications.

[0074] Therefore, it is necessary to correct the maturity index of the fruit to be identified based on the difference index to obtain the corrected maturity index value of the fruit to be identified. The formula used is as follows:

[0075]

[0076] in, This represents the maturity index correction value of the fruit to be identified. The maturity index of unripe fruit; in the above correction formula, A smaller value indicates a greater difference between the fruit to be identified and a mature fruit, suggesting lower maturity. A larger value indicates less difference and higher maturity.

[0077] The maturity index of the fruit to be evaluated is obtained through neural network prediction. This represents a preliminary assessment of the fruit's maturity. This value reflects the initial results obtained by the network model through analyzing the fruit image. When the difference index is small, it indicates that the fruit to be identified is close to ripeness. Therefore, the corrected maturity index... It will be close to or greater than the basic maturity index. A higher maturity index indicates a higher degree of fruit maturity, leading to more accurate system predictions. A larger difference index indicates that the fruit to be identified is not yet close to maturity; therefore, the corrected maturity index... This will decrease, and the maturity level predicted by the system will be revised to be more conservative.

[0078] It should be noted that this logical formula effectively determines whether fruit has reached the harvesting standard by comparing the modified maturity index with a set maturity threshold. This judgment process is crucial because it ensures that only ripe fruit can be harvested, thus avoiding waste and quality degradation caused by picking unripe fruit. Threshold T T The settings need to be adjusted according to the actual situation, and are usually determined based on factors such as fruit variety, climate conditions, and ripening cycle, to ensure that the prediction system can accurately determine the optimal harvesting time. This approach improves harvesting efficiency, reduces resource waste, and ensures the accuracy and economy of harvesting operations.

[0079] Therefore, it is necessary to compare the maturity index correction value of the fruit to be identified with the maturity threshold to determine whether the target fruit can be harvested. The logical formula used is as follows:

[0080]

[0081] in, This indicates the established maturity threshold. The logical value used to determine whether the target fruit can be harvested is when... When the maturity index correction value of the fruit to be identified is less than the maturity threshold, it indicates that the fruit is not yet ripe and cannot be harvested; when When the maturity index correction value of the fruit to be identified is greater than or equal to the maturity threshold, it indicates that the fruit to be identified is mature and harvesting should be carried out immediately.

[0082] Please see Figure 2 The present invention also provides a fruit picking system based on target recognition, the fruit picking system being used to perform the above-described fruit picking method based on target recognition, comprising:

[0083] The feature data acquisition module is used to acquire image data of the fruit to be identified and to preprocess the image of the fruit to be identified, including noise reduction and contrast enhancement.

[0084] The data preprocessing module is used to construct a fruit maturity prediction model using a convolutional neural network. The fruit maturity prediction model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive preprocessed image data, the hidden layer is used to process the image data, and the output layer is used to output the fruit maturity index. The image data of the fruit to be identified is input into the prediction model to obtain the maturity index of the fruit to be identified.

[0085] The picking model establishment module is used to extract feature data of the fruit to be identified from the image of the fruit to be identified. The feature data includes roundness, contrast, energy and entropy. It also obtains feature data of the fruit maturity stage, calculates the difference value between the maturity stage and the feature data of the image of the fruit to be identified, generates a difference index using the image feature difference value, and corrects the maturity index of the fruit to be identified based on the difference index to obtain the corrected maturity index value of the fruit to be identified.

[0086] The threshold determination module is used to set a maturity threshold, compare the maturity index correction value of the fruit to be identified with the maturity threshold, and determine whether the target fruit can be harvested.

[0087] This invention provides a computer device, comprising:

[0088] The electronic device includes a memory for storing a computer program and a processor for running the computer program to enable the electronic device to perform the above-described fruit-picking method based on target recognition.

[0089] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer, when executed by a processor, implements the above-described fruit picking method based on target recognition.

[0090] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0091] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0093] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A fruit picking method based on target recognition, characterized in that, The specific steps include: Step 1: Collect image data of the fruit to be identified and preprocess the image of the fruit to be identified. The image preprocessing includes noise reduction and contrast enhancement. Step 2: Construct a fruit maturity prediction model using a convolutional neural network. The fruit maturity prediction model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive preprocessed image data, the hidden layer is used to process the image data, and the output layer is used to output the fruit maturity index. The image data of the fruit to be identified is input into the prediction model to obtain the maturity index of the fruit to be identified. Step 3: Extract the feature data of the fruit to be identified from the image of the fruit to be identified. The feature data includes roundness, contrast, energy and entropy. Also, obtain the feature data of the fruit's ripening stage. Calculate the difference value between the ripening stage and the feature data of the image of the fruit to be identified. Use the image feature difference value to generate a difference index. Based on the difference index, correct the ripening index of the fruit to be identified to obtain the corrected ripening index value of the fruit to be identified. Step 4: Set a maturity threshold, compare the maturity index correction value of the fruit to be identified with the maturity threshold, and determine whether the target fruit can be harvested. The difference between the image feature data of the ripening stage and the current image of the fruit to be identified is calculated. A difference index is then generated using this difference value, based on the following formula: in, , , These represent the differences in roundness, contrast, energy, and entropy of the current fruit image feature data to be identified. , , , The roundness, contrast, energy, and entropy of the collected fruit image feature data at the ripening stage are analyzed. , , The features of the fruit image to be identified include roundness, contrast, energy, and entropy. The formula used to generate the difference index from image feature differences is as follows: in, Indicates the degree of difference; The maturity index of the fruit to be identified is corrected based on the difference index to obtain the corrected maturity index value of the fruit to be identified. The formula used is as follows: in, This represents the maturity index correction value of the fruit to be identified. This is the fruit maturity index.

2. The fruit picking method based on target recognition according to claim 1, characterized in that, Image preprocessing includes denoising and contrast enhancement, while image data preprocessing includes data cleaning and normalization. The methods used are as follows: Denoising preprocessing of image data involves using filtering techniques to reduce noise in the image. While preserving image information, noise removal ensures the accuracy of subsequent feature extraction. Secondly, contrast enhancement is performed through adaptive contrast enhancement, which enhances the contrast of local areas to better preserve image details, thereby improving the accuracy of the fruit recognition and picking model.

3. The fruit picking method based on target recognition according to claim 1, characterized in that, A fruit maturity prediction model was constructed to obtain the maturity index of the fruit to be identified. The method used was as follows: The fruit maturity prediction model is trained by acquiring a historical fruit image dataset and determining the fruit maturity index based on expert scores. The mean squared error function is selected as the loss function. The loss function value is calculated based on the output results and the true labels. The gradient is calculated through the backpropagation algorithm, and the weights and biases of the neural network are updated. The above operations are repeated until the model reaches the predetermined number of training rounds.

4. The fruit picking method based on target recognition according to claim 1, characterized in that, The corrected maturity index value of the fruit to be identified is compared with the maturity threshold to determine whether the target fruit can be harvested. The logical formula used is as follows: in, This indicates the established maturity threshold. The logical value used to determine whether the target fruit can be harvested is when... When the maturity index correction value of the fruit to be identified is less than the maturity threshold, it indicates that the fruit is not yet ripe and cannot be harvested; when When the maturity index correction value of the fruit to be identified is greater than or equal to the maturity threshold, it indicates that the fruit to be identified is mature and harvesting should be carried out immediately.

5. A fruit-harvesting system based on target recognition, characterized in that, The fruit harvesting system is used to execute the fruit harvesting method based on target recognition as described in any one of claims 1-4, including: The feature data acquisition module is used to acquire image data of the fruit to be identified and to preprocess the image of the fruit to be identified, including noise reduction and contrast enhancement. The data preprocessing module is used to construct a fruit maturity prediction model using a convolutional neural network. The fruit maturity prediction model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive preprocessed image data, the hidden layer is used to process the image data, and the output layer is used to output the fruit maturity index. The image data of the fruit to be identified is input into the prediction model to obtain the maturity index of the fruit to be identified. The picking model establishment module is used to extract feature data of the fruit to be identified from the image of the fruit to be identified. The feature data includes roundness, contrast, energy and entropy. It also obtains feature data of the fruit maturity stage, calculates the difference value between the maturity stage and the feature data of the image of the fruit to be identified, generates a difference index using the image feature difference value, and corrects the maturity index of the fruit to be identified based on the difference index to obtain the corrected maturity index value of the fruit to be identified. The threshold determination module is used to set a maturity threshold, compare the maturity index correction value of the fruit to be identified with the maturity threshold, and determine whether the target fruit can be harvested.

6. A computer device, characterized in that, include: A memory and a processor, the memory for storing a computer program, the processor for running the computer program to cause the computer device to perform the fruit picking method based on target recognition as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer is executed by the processor, it implements the fruit picking method based on target recognition as described in any one of claims 1-4.

Images