Hi,

Your question is a bit ambiguous, as it very much depends on how the fruit is described (images, sensor measurements, amount of physical components, etc...) by features. So if you want a specific answer, you should give more precisions.

If you happen to have images, access to sufficient computation power, and if accuracy is the only goal you are aiming it, then most probably deep learning techniques can do marvels for you, especially as fruit images can be collected easily on the web (so lots of training data).

For example, such techniques are used in the Plantnet project (http://www.plantnet-project.org/papyrus.php?langue=en) to recognize plants use such techniques.

Best regards

Hello

Its totally depends on the data you are working with. The feature extraction is the very important step before applying classification. The classification accuracy is not only dependent on the classifier but also on the features you select. The separability (using mean and standard deviation can be calculated to select best feature set. You can use SVM/ deep learning or random forest (non parametric classifiers).

It’s depending on the application data you are working with. If your application involve with the image processing technics you can use multiple features based algorithm (Intensity and Color Features extraction, Orientation Features extraction, Edge Features extraction, etc.). To detect the fruit defects use k-means clustering and graph-based algorithms.

The accuracy of any pattern classification problem depends on lot of parameters. So, there is no so-called 'best' feature or classifier applicable for all. But you can try some texture based feature (or you customize the same to fit it into your data set).

Fruit classification using image processing has many considerations (I assume that is 1. your question). Nice literature is available, please see some of the articles published:

1. Date fruits classification using texture descriptors and shape-size features 

2. Application of Image Processing in Fruit and Vegetable Analysis: A Review 

3. Fruit classification using computer vision and feedforward neural network 

4. Automatic fruit and vegetable classification from images 

5. Fruit and vegetable recognition by fusing colour and texture features of the image using machine learning 

6. Classification of Fruits Using Computer Vision and a Multiclass Support Vector Machine 

7. Automatic fruit classification using random forest algorithm 

8. A Code based Fruit Recognition Method Via Image Convertion Using Multiple Features 

9. Study on Fruit Recognition Methods Based on Compressed Sensing 

The problem of fruit classification starts with types of fruits (as more number of fruits, larger the complexity), types of input image (RGB, Ycbcr or hyper spectral) which again is related to your sensor (fruit capturing device).

First step involves image preprocessing for extraction of the fruit from the captured image and color space conversion. 

Prime step is the feature extraction step. Mostly all the algorithms uses various color, texture and shape descriptors for feature extraction. You can refer above listed articles for the same. Another important step involves feature preprocessing (feature selection) which reduces the feature dimensionally and sometimes classification accuracy too. ANN and SVM are two popular choice for the last classification step. But you can choose any other classifiers too to check their performance.