python

The identification and diagnosis of plant diseases have long been considered. This research presents a system for diagnosing the volume and type of apple diseases and the spoilage percentage of rotten apples. To estimate the volume of apples, the method of immersion in water to change the volume of the container was used, ensuring more accurate volume estimation. For disease detection and spoilage analysis, a chamber with constant lighting conditions and a halogen lamp was used. Four images were taken with a camera for better analysis. The volume of apples was calculated through two approximations of the cylinder and incomplete cone. The average error rate in this system was 5%. Also, in the present research, a novel method for feature selection was identified using a combination of the weight feature and the calculated volume of hollow apples. To calculate the percentage of failure of each apple, first, the type of failure was identified. Then, the ratio of loss of each apple relative to the whole apple was calculated and compared with the number obtained from the desired region method, which was accurate. In this study, three major diseases of apples were studied, and an algorithm was written to distinguish these three types of infections from healthy apples. The results showed that the proposed method had the necessary efficiency to calculate the volume and percentage of failure and diagnose the type of apple diseases. In addition, the system's accuracy compared to previous studies increased by up to 95%.

A deep learning-based convolutional artificial neural networks structured a new image classification method approach was implemented in the study. Sample application was carried out with Diabetic Retinopathy disease. Obtaining information about the blood vessels and any abnormal patterns from the rest of the phonoscopic image and assessing the degree of retinopathy is the problem itself. To solve this problem developed methodology and algorithmic structure of this new approach is presented in the study. An approach called care model was used in this study different from the classical CNN structure. The care approach is based on the idea that the best solution will be taken from the new data obtained by rescale the available data according to total number of pixels before the average data pool is created and then CNN processes will continue. In the care model approach, all data is multiplied by the number of elements by the number of epoch time eight tensors. The purposed care model include VGG19 image classification model and developed mathematical model presented. Pre-trained model and all image dataset taken from kaggle and keras for implementation of case study. The purposed model provide train accuracy 87%, test accuracy 88%, precision 93% and recall 83%.