Sensor Based Sorting can decrease the amount of material processed in the subsequent processing steps, by separating the valuable material from the invaluable material, and therefore potentially reducing costs in those subsequent steps. For sensor based sorting to be effective, a strong relation needs to be present between the properties detectable with real-time sensors and the valuable content. This thesis report describes a preliminary study, which aims to evaluate the potential of the use of RGB images in predicting gold presence and abundance in drillhole samples from the Cortez Hills underground mine in Nevada, U.S.A. The goal of this research is to find a relation between image data and geochemical data. The image data include RGB images acquired by a Specim core logger of 628 drillcore samples from the Cortez Hills Lower Zone deposit. Material from the same drillcores has been analyzed geochemically by Barrick. The results of the geochemical analysis provide information which include the gold grade, gold recovery and arsenic content. This information is taken at approximately the same depths as the image samples are retrieved from, creating 628 geochemical measurement locations. The research focuses on comparing the image data and the corresponding geochemical data. In order to understand the geological history of the deposit, literature was used to describe and assess the deposit. The deposit at Cortez is a Carlin-Type deposit. Carlin-Type deposits are characterized by extremely fine-grained disseminated gold hosted primarily by arsenian pyrite. Carlin type deposits are dated as Eocene age, more specifically, the mineralization at Cortez is dated at 34 Ma. The Cortez Hills deposit consists of disseminated gold particles in highly altered calcareous host rocks. Because of the the complexity of its geological history, the appearance of the samples varies and this variation is possibly linked with the gold mineralization.The geochemical data is used to replicate the classification of ore types known at Cortez: Refractory ore (Autoclave ore & Roaster ore), Oxide ore (Mill Oxide ore & Leach Oxide ore) and Waste. The Refractory ore is defined as material with a gold grade higher than 4.69 ppm and a gold recovery lower than 0.7. Autoclave ore has an arsenic content of more than 1200 ppm, and below that value the refractory ore is called Roaster ore. Oxide ore is defined as material with a gold grade higher than 0.125 ppm and a gold recovery higher than 0.7. Mill oxide ore is high grade ore (>4.69 ppm Au) and Leach oxide ore is medium grade ore (<4.69 ppm Au). The other material is called waste. This research aims to distinguish these 5 ore types (Autoclave, Roaster, Mill, Leach and Waste) by the image data. The image data was analyzed and compared to these ore types using color based parameters and textural parameters both defined in Matlab. 19 Color based parameters were created based on RGB and HSV values. 28 Textural parameters are obtained using functions from the image Processing Toolbox of Matlab. All parameters are compared with the different ore types. Of all color based and textural parameters, the average blue intensity parameter is the best for separating ore and waste: Half of the samples below a blue intensity of 0.3 are classified as ore, and above this threshold 89.7% was classified as waste. Another parameter was defined as the percentage of pixels with a red intensity 0.05 higher than both the green and blue intensity. This parameter distinguishes the oxide ore samples, since they have a higher percentage for the parameter than the refractory ore samples. 7 Color based parameters were used to create a classification tree model. This model classifies the samples into 8 different visual appearances: Black, Dark, Dark&White, White, Red, Dark&Red, Pink and Grey. These classes are compared to the different ore types, in order to investigate the relations between the appearance classes and the ore types. The results of this model indicate that the classes have relations with the ore types: 63%of the classes Red and Dark&Red are oxide ore samples (Leach and Mill), 82% of the samples with a Black class are Refractory ore samples (Autoclave and Roaster) and the classes Dark&White, Pink, Grey and White comprise 54% of the data set and consist for 86.7% out of Waste.The color based parameters and textural parameters are also put in one matrix making a data matrix used for PCA and PLS-DA models. The scores and responses of these models are again compared to the 5 ore types. PLS-DA proved to be the best method in separating oxide ore, refractory ore and waste based on a set of 47 image data variables. By combining multiple PLS-DA models, 87.2% of the data set is classified in three groups, Refractory, Oxide and Waste, with an accuracy of 75.8%. The research showed relations between RGB image data and both gold presence and gold recovery and therefore as well between RGB image data and the types refractory, oxide and waste. In ore sorting applications, this could not only mean a significant decrease in the volume of the feed, but also a valuable addition to blending purposes. Because of the relations found between image data and gold content, the Cortez Hills underground mine and other Carlin-type mines in the area may show potential for implementation of sensor based sorting in the processing operation at the mine.