INVESTIGATION OF RUT PATTERNS IN DIFFERENT TERRAIN CONDITIONS FOR VEHICULAR MOVEMENTS

Abstract

The movement of vehicles on an unpaved terrain is a common requirement in many fields including agriculture, forestry, automobile, planetary rovers and defence. Other than surface topographical features, the most important parameter to the movement of vehicles is the underlying soil condition. Acquiring the data by conventional methods is laborious and cumbersome task. The alternative means have therefore been explored by various researchers. One of the most effective ways employed for anlysing the prevalent soil condition is by monitoring the rut formed by vehicular movement in the area. Many soil parameters like soil condition, its gradation, moisture content, soil strength etc. impact the vehicular rut. Vehicle loading conditions like tyre size, vehicle weight, its speed, curvatures, and repeated passes also influence the rut shapes. Although many parametric studies have been conducted to characterize and model the rut shapes based on all these, yet several aspects are still to be studied. One aspect of the issues pertains to modelling and evaluation of rut depth. Most of the literature focuses on evaluating the rut depth. Certain issues are however typical in different scenarios which need to be addressed. The rut in desertic terrain has been observed to get filled by the sand pouring from sides. Similarly, the rut profile has been observed to become eccentric on the curves. This aspect demands for mapping the shapes of rut profiles in different terrain-vehicle running conditions. Moreover, with advent of technology, the rut profile measurement tools too have moved from manual to advanced laser-based sensors. The laser profiler on one hand measures the rut profile with precision, however, it needs heavy memory devices for storage and interpretation of data. The optimization aspect of rut profile data needs to be explored for efficient movement decisions. Further, a number of models are developed that try to characterize the influence of different causative factors on rut. While selection of appropriate model is one aspect, the overall impact of all causative factors on the maximum soil distress levels in any area is important to be studied for ascertaining the suitability of the terrain for any emergency movement. Another aspect of rutting research pertains to addressing the issues in wider spatial domain. While evolving suitable spatial models governing trafficability potential is an important aspect, the validation of interpreted information is another important area needing attention. Here, identification of track impressions that look like edges in coarse resolution images can provide useful information about identifying the trafficable zones. Identifying the tracks manually being tedious, alternate means need to be explored. Moreover, the rut tracks formed by the leading vehicle is said to provide useful information for the rut following robotic vehicles, v defence, and forestry. The delineation of track impressions form surrounding terrain and visibility conditions is an important consideration needing attention. In this research work, rut has been investigated from different perspectives. One aspect focuses on the experimental studies of rut profiles in different fields while the other one tries to address the issue of delineation of rut tracks by collating various image processing techniques. In the field based experimental studies, various shapes of rut profiles on different types of soil and vehicle running conditions were investigated. The rut profile data was captured using both manual and laser-based systems. The most common rut shapes observed in field are identified and grouped in different categories. Attempt was then made to devise better ways for optimal storage of most common rut profiles. By using the proposed mathematical formulations, an additional compression of more than 80% over the conventional compression techniques could be achieved on straight patches and 71% on turnings. In another experimental study, soil distress level was investigated using multiple vehicular passes on varied terrain conditions. This study paves the way for identifying and mapping the unpaved areas suitable for planning emergency support. Another part of study focused on visual enhancement and detection of rut-based track impressions. In order to detect edges like track features in satellite images, various edge detection algorithms are explored. The comparative study of different algorithms revealed that the Canny Edge detection method gives relatively better results. Further studies are however needed for improved detection and delineation of tracks. It is observed that the tracks formed by vehicular rut impressions appear like thin edges in images of coarse-resolution. However, when using the fine resolution images, the same features appear like elongated areas. In such a scenario, the edge detection filters and even the conventional contrast enhancement techniques are able to delineate these features to a limited extant. The role of texture of these tracks that can differentiate these tracks from their surroundings has been explored in this study. Gray level co-occurrence matrix (GLCM) which is a texture measurement technique has been employed here. To compare the effectiveness of various techniques in enhancement of track contrast in a given surrounding, a new quantitative track index (TI) based measure has been proposed in this study. Here, the effectiveness of technique in enhancing the track contrast has been evaluated. Various forms of track indices as proposed in this study have been compared. The proposed track index effectively sorts correctly the contrast images to the level of 88%. The proposed track index-based technique is seen as effective means for sorting the images based on track contrast. This method can bring in improved fidelity of decisions for the sustainable operations. The study was extended further and a new technique based on track index has been developed that is seen as adaptive for enhancing the track contrast in a given surrounding. The outcome of above research has been presented in various chapters of this thesis. The approach of bringing in optimization in data storage is a step towards making efficient decisions about trafficability condition of the terrain. The evaluation of maximum soil distress level under different dynamic conditions sets another way of identifying and mapping the safe trafficable zones for planning emergency vi movement. The image analysis-based improved identification of rut tracks is an important contribution in visual analytics-based systems on-board vehicles. The mobility decisions could be made better and efficient using this track index based technique. The edge detection algorithm could set the way for improved identification of unpaved tracks in satellite images. Further research is however needed for automated delineation of rut tracks for inferring trafficable zones. The machine learning approach could be explored here.