1. A computer implemented method for identifying and quantifying sonar targets within a liquid medium, the method comprising the steps of:
collecting a raw sidescan sonar image;
separating a region of interest related to the sonar targets from the image;
performing an image transformation on the image using an extraction algorithm;
performing particle analysis on the extracted region of interest to generate a feature vector related to sonar targets; and
presenting the generated feature vector to a neural network to classify the image with respect to the sonar targets of interest.
2. A method according to claim 1, wherein the neural network is a radial basis function neural network.
3. A method according to claim 2, wherein the radial basis function neural network is trained with prototype sidescan sonar images by performing the steps of:
obtaining a feature vector derived from the prototype image;
presenting the image vector to the radial basis function neural network;
if the presented feature vector is not within an influence field of any prototypes already stored in the network, committing a new neuron to the presented vector;
if the presented feature vector falls within an influence field of an already learned vector in an existing neuron, making no change to the radial basis function network; and
if the presented feature vector fall within a wrong influence field or is mismatched to a category, readjusting one or more influence fields.
4. A method according to claim 1, wherein the performing an image transformation step further comprises:
calculating length, width, area and mean pixel intensity values;
applying a threshold operator; and
removing spurious pixels from the image to obtain an extracted region of interest.
5. A method according to claim 4, wherein the spurious pixels are removed by a dilation operation or an erosion operation.
6. A method according to claim 1, wherein the presenting a feature vector step further comprises:
determining whether the feature vector lies within an active influence field of a prototype in the neural network;
if the vector is not within the active influence field of any prototype in the neural network, classifying the feature vector as not recognized; and
if the feature vector is within the active influence field of any prototype in the neural network, recognizing the input as belonging to the active influence field’s corresponding category.
7. A method according to claim 1, wherein the raw sidescan sonar image is collected by a digital signal processor.
8. A method according to claim 7, wherein the sidescan sonar image is based on a frequency of about 600 kHz.
9. A method according to claim 1, wherein the sidescan sonar image is based on a frequency ranging from about 100 kHz to about 2.4 MHz.
10. A method according to claim 1, wherein the region of interest is separated from the image by an extraction algorithm.
11. A method according to claim 1, wherein the raw sidescan sonar image is collected by an autonomous underwater vehicle.
12. A system for identifying and quantifying sonar targets of interest within a liquid medium comprising:
an autonomous underwater vehicle;
a transducer mounted on the autonomous underwater vehicle to generate a sidescan sonar image;
a processor, for collecting the sidescan sonar image, housed inside the autonomous water vehicle, the processor configured to:
separate a region of interest related to the sonar targets from the image;
perform an image transformation on the image using an extraction algorithm;
perform particle analysis on the extracted region of interest to generate a feature vector related to sonar targets; and
present the feature vector to a neural network to classify the image with respect to the sonar targets of interest.
13. The system of claim 12 further comprising a communication unit housed in the autonomous underwater vehicle for automatic reporting of positive identification of the sonar targets of interest.
14. The system of claim 12, wherein the perform an image transformation by the processor step further comprises:
calculating length, width, area and mean pixel intensity values;
applying a threshold operator; and
removing spurious pixels from the image to obtain an extracted region of interest.
15. The system of claim 12, wherein the neural network to classify the image comprises a radial basis function neural network.
16. The system of claim 15, wherein the radial basis function neural network is trained with prototype sidescan sonar images by performing the steps of:
obtaining a feature vector derived from the prototype image;
presenting the feature vector to the radial basis function neural network;
if the presented feature vector is not within an influence field of any prototypes already stored in the network, committing a new neuron to the presented vector;
if the presented feature vector falls within an influence field of an already learned vector in an existing neuron, making no change to the radial basis function network; and
if the presented feature vector fall within a wrong influence field or is mismatched to a category, readjusting one or more influence fields.
17. The system of claim 12, wherein the processor comprises a digital signal processor for collecting the sidescan sonar image.
18. The system of claim 12, wherein the transducer has a range setting in the range of five to ten meters.
19. The system of claim 12, wherein the present a feature vector step in the processor further comprises:
determining whether the feature vector lies within an active influence field of a prototype in the neural network;
if the vector is not within the active influence field of any prototype in the neural network, classifying the feature vector as not recognized; and
if the feature vector is within the active influence field of any prototype in the neural network, recognizing the input as belonging to the active influence field’s corresponding category.
20. A computer readable medium having program code recorded thereon, that when executed on a processor, identifies and quantifies a sonar target of interest in a liquid medium, the program code comprising:
code for receiving a sidescan sonar image from a sonar region being monitored;
code for separating a region of interest related to the sonar targets from the image;
code for performing an image transformation on the image using an extraction algorithm;
code for performing particle analysis on the extracted region of interest to generate a feature vector; and
code for presenting a feature vector related to the sonar targets to a neural network to classify the image with respect to the sonar targets of interest.
The claims below are in addition to those above.
All refrences to claim(s) which appear below refer to the numbering after this setence.
1. A continuous process for drying a material containing an initial degree of water which comprises the steps of:
(a) driving a vehicle comprising a multi-mode microwave applicator over at least a portion of said material;
(b) exposing said portion of said material to said applicator having an air flow about said portion of said material;
(c) exposing said portion of said material to at least two sources of microwaves, said microwaves being in non-parallel alignment to each other for a period of time sufficient to dry said portion of said material to a lower degree of water;
said at least two sources of microwaves propagated from a bifurcated waveguide assembly;
said introduced microwaves, being 90\xb0 out of phase to each other;
said microwaves having a frequency between 915 MHz and 1000 MHz;
said microwave applicator further comprising a 3-port ferrite circulator to absorb any reflected microwaves; and
said bifurcated waveguide assembly having a microwave pressure window at an exit end comprising fused quartz to prevent any vapors in the applicator from returning through said waveguide.
2. The process of claim 1 wherein said applicator is heated and operates between approximately 100\xb0 F. and 212\xb0 F.
3. The process of claim 2 wherein said applicator is heated by a heating means provided from combustion products of engine exhaust gas ducted into and out from the microwave applicator.
4. The process of claim 1 said frequency is approximately 915 MHz.
5. The process of claim 1 wherein said material is a roadbed.
6. The process of claim 1 wherein said material is concrete.
7. The process of claim 1 wherein material is an asphalt surface.
8. The process of claim 1 wherein said material is an agricultural field.
9. A drivable apparatus which comprises:
(a) a movable chassis comprising a microwave generator;
(b) at least one multi-mode microwave applicator in proximity to a material to be dried and in communication with said microwave generator via a waveguide;
(c) said applicator having an air flow about at least a portion of said material;
(d) said applicator having at least two sources of microwaves from said microwave generator, said microwaves being in non-parallel alignment to each other;
said at least two sources of microwaves propagated from a bifurcated waveguide assembly;
said introduced microwaves, being 90\xb0 out of phase to each other;
said microwaves having a frequency between 915 MHz and 1000 MHz; and
(e) a microwave energy absorber to absorb any reflected microwaves; and
(f) said bifurcated waveguide assembly having a microwave pressure window at an exit end comprising fused quartz to prevent any vapors in the applicator from returning through said waveguide.
10. The apparatus of claim 9 wherein said applicator is heated and operates between approximately 100\xb0 F. and 212\xb0 F.
11. The apparatus of claim 10 wherein said heated applicator is heated by a heating means provided from combustion products of an engine exhaust gas ducted into and out from the microwave applicator.
12. The apparatus of claim 9 said frequency is approximately 915 MHz.
13. The apparatus of claim 9 wherein said microwave energy absorber is a 3-port ferrite circulator.
14. The apparatus of claim 9 which further comprises an RF trap.