Research On Extraction And Identification Of Common Wild Edible Fungi

Nowadays,image recognition has been widely used in the field of agricultural engineering to detect state of crop growth,pest and disease prediction,and safety diagnosis.Feature extraction algorithm,as an important binding in the entire recognition process,affects the quality of the final classification result.Feature extraction targets mainly focus on the shape,color and texture of agricultural products.The automatic picking of wild edible fungi needs to provide identification technology,and the subsequent processing must be effectively classified,the prerequisite for achieving these goals is to extract the feature of edible fungi.This paper proposes an improved feature extraction method based on common wild edible fungus images,which effectively promotes the accuracy of image recognition.1.This study uses the common wild edible fungi in the Greater Khingan Mountains Prefecture as object.First of all,I collected more than 500 edible fungus images from the natural environment,establish a database of wild edible fungi,and used MATLAB and image processing methods to study the image acquisition process,image processing process,and image recognition process images.2.The purpose of image processing was to effectively extract the characteristic parameters and improve the accuracy and speed of detection of edible fungi.In this research,the collected edible fungus images were effectively denoisied,the target image was segmented,and the mathematical morphological processing were performed to provide clear edges for feature extraction of wild edible fungi.3.The method of extracting the morphological features of the HSV color space quantization scheme for edible fungi was studied.By comparing different color spaces,an improved color quantization scheme(16: 3: 3)was proposed,and H,S,and V color component histograms are extracted,and then combined with shape feature to identify fungi using Otsu algorithm.Indexes of comprehensive evaluation were used to quantitatively evaluate the experimental results.The results were evaluated by indexes,showing that the edible fungus recognition rate reached 91.33%,and the average time to identify an image was 3.28 s.4.The edible mushroom fold texture feature extraction method was studied,and a combination of discrete wavelet transform,local three-valued mode,and rotation-invariant gray level co-occurrence matrix was proposed to extract 12 valid feature value parameters.FCM clustering and support vector machine algorithm were used to classify and identify similar edible fungi,and compared with the present extraction methods,99.7% classification accuracy was obtained.Through the comparative analysis of inter-class evaluation methods,it is concluded that this method can effectively classify and identify wild edible fungi.The study has capability to is significant for effectively identifying and classifying similar edible fungi in the wild,which provides a technical support for the picking automation and the trade export of wild edible fungi.