Research Of Strawberry Maturity Detection Based On Deep Learning

As an important fruit crop,strawberries are not only delicious but also rich in nutrients.However,there are many problems in the growth of strawberries,such as short picking cycle and difficult storage of fruits,which bring challenges to the cultivation and picking of strawberries.Therefore,the realization of fast and accurate strawberry detection and location positioning to assist agricultural producers to pick ripe strawberries in time can effectively improve picking efficiency and reduce picking costs,and the realization of strawberry object detection has very important practical significance and broad application potential.At present,there is no large-scale data set for the strawberry detection task.This research mainly uses the self-shooting method to obtain the original strawberry image,and performs data enhancement by means of rotation,color transformation and adding noise to simulate the strawberry growth environment in the natural environment.Through The data samples were marked,and the strawberry data set was successfully constructed,which provided rich sample data for subsequent research.In order to further improve the detection effect of strawberry objects,this paper proposes a object detection model that integrates residual network and attention mechanism based on deep learning technology.The model replaces the feature extraction network with a residual network structure,adopts a multi-scale feature extraction method,uses different sizes of prior frames to traverse the feature maps of different levels,and uses the attention mechanism method to more comprehensively capture the strawberry image.Feature information,using the combination of position positioning error and classification error to define the loss of the algorithm model.The experimental results show that the improved model has a detection accuracy of 98.4% for strawberries,99.0% for mature strawberries,and 97.7% for immature strawberries,effectively improving the ability to identify the maturity of strawberries.It provides an experimental basis for the development of strawberry maturity research.In order to realize the mobile terminal detection of strawberry objects,this paper constructs a strawberry lightweight object detection model.The model uses a lightweight network structure for information feature extraction,and also applies a lightweight attention mechanism to further improve detection performance.At the same time,through the improved loss function Speed up the model convergence process and make the training process more efficient.The improved lightweight model has a transmission rate of 134 FPS for strawberry images with a size of 320×320,and can maintain a high detection speed while maintaining detection accuracy.It has better adaptability to mobile terminals with limited computing resources..Based on the lightweight object detection model,this research uses the Python programming language and the Py Qt5 extension package to develop a strawberry object detection system.The system has the ability to perform real-time detection in various scenarios,and can detect strawberry images and video files.,and can call the camera to complete real-time detection.In addition,the lightweight model was deployed to the Android mobile phone,and the model was deployed using the NCNN framework and Android Studio tools,and the strawberry image detection on the mobile phone was completed.Mobile terminal detection provides a feasible solution for strawberry detection in the natural environment,and also provides better technical support for agricultural production.